[
 {
  "id": 4306,
  "day": "2026-12-01",
  "type": "Opening remarks",
  "start": "09:00",
  "end": "09:15",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Welcome and Opening Remarks",
  "speakers": [
   {
    "name": "Mehdi Medjaoui",
    "title": "Author of Continuous API Management",
    "company": "Founder and Chairman of Apidays Conferences"
   },
   {
    "name": "Margaux Pelen",
    "title": "founder",
    "company": "episcope"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Opening remarks, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4783,
  "day": "2026-12-01",
  "type": "Keynote",
  "start": "09:15",
  "end": "09:35",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Simon Frieder",
    "title": "Researcher",
    "company": "University Oxford"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4298,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "09:15",
  "end": "09:55",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Why API Management Isn't Enough",
  "speakers": [
   {
    "name": "Marjukka Niinioja",
    "title": "API Consultant & APIOps Community lead",
    "company": "Osaango Ltd"
   }
  ],
  "facets": [
   "contract_governance",
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Operating-model failures after API management: governance, adoption and evidence, not a published surface.",
  "abstract": "Why platform technology is not the hardest part anymore What happens Frame \"After API Management\" Name common operating-model failure modes Connect platform investment to adoption, governance, AI-speed, and evidence Explain how the participant-shaped day will work Outcome Shared vocabulary Common problem statement Understanding of how to participate in the day",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4784,
  "day": "2026-12-01",
  "type": "Keynote",
  "start": "09:35",
  "end": "09:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Decoding the Language of Life: When Natural Language Meets DNA",
  "speakers": [
   {
    "name": "Loubna Ben Allal",
    "title": "Research Lead",
    "company": "HuggingFace"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; DNA and natural language models, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4785,
  "day": "2026-12-01",
  "type": "Keynote",
  "start": "09:55",
  "end": "10:15",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "AI For Physics Discoveries",
  "speakers": [
   {
    "name": "David Louapre",
    "title": "Science Etonnante, Researcher",
    "company": "Hugging Face"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; AI for physics research, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4916,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "09:55",
  "end": "10:15",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "MCP Needs More Domain-Specific Extensions",
  "speakers": [
   {
    "name": "Yann Jouanin",
    "title": "CEO",
    "company": "MCP APPS BUILDERS"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "mcp_tool_quality"
  ],
  "why": "No abstract; domain-specific MCP extensions are shared vocabulary for agent tool surfaces.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4299,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "09:55",
  "end": "10:40",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Real-World API & Integration Challenges",
  "speakers": [
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   },
   {
    "name": "Claire Barrett",
    "title": "Founder",
    "company": "APIsFirst"
   },
   {
    "name": "Miriam Greis",
    "title": "Principal Integration Architect.",
    "company": "adorsys GmbH + Co. KG"
   },
   {
    "name": "Miguel Quintero",
    "title": "Technical Trainer",
    "company": "Postman"
   }
  ],
  "facets": [
   "discoverability",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [],
  "why": "Why we still don't know what APIs we have, and why operating-model change stalls.",
  "abstract": "Real stories, participant challenges, and a vote on what the room should work on. Speaker topics Kin Lane: Why we still don't know what APIs we have Claire Barrett: Why operating-model change gets stuck Miriam Greis: Practitioner spark to be confirmed Miguel Quintero: Practitioner spark to be confirmed What happens Hear four short practitioner provocations: what did not work, what changed, or what remains difficult Add participant-submitted challenges to the marketplace Vote on which station clinic priorities deserve the room's attention Outcome Clinic priorities for the day Sharper view of the problems represented in the room",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4917,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "Agents Without Users: Giving Every Agent Its Own Identity",
  "speakers": [
   {
    "name": "Aaron Parecki",
    "title": "Sr Director of Identity Standards",
    "company": "Okta"
   }
  ],
  "facets": [
   "agent_readiness",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [
   "agent_identity_declared",
   "delegated_identity",
   "dynamic_client_registration"
  ],
  "why": "No abstract; a distinct identity for every agent.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4957,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Thomas Roux",
    "title": "AI Productivity Lead",
    "company": "Sanofi"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4811,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "Text and tool calls in one token stream: a hidden hard part of agents",
  "speakers": [
   {
    "name": "Jérémie Doucy",
    "title": "CTO",
    "company": "Bayes Impact"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality"
  ],
  "why": "How models switch to tool calls mid-stream and where that breaks; client-side, little provider-surface signal.",
  "abstract": "Tool calling looks solved: send tools, get JSON back. Underneath, the model streams tokens and switches to a tool call mid-stream, and each model marks that switch differently. Based on agents we run in production, this talk shows where tool calling breaks, how to test it and how to fix it without touching the model’s weights.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4933,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Sovereignty Through Openness, from Domyn Edge 4B to EUROPA 1T",
  "speakers": [
   {
    "name": "Luca Antiga",
    "title": "CTO",
    "company": "Domyn"
   }
  ],
  "facets": [
   "open_source",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Open models as a path to sovereignty; no facet reads model sovereignty.",
  "abstract": "Sovereignty is ownership and control. At Domyn we argue for openness as the path to sovereignty, for Europe and for individual organizations. An ecosystem building on shared open foundations can achieve what no single entity can alone, while fully controlling the stack and the playbooks. This talk presents our contributions to that ecosystem. Domyn Edge, state-of-the-art 4 B‑parameter model trained from scratch on European public compute, and fully open (weights, intermediate checkpoints, all data, code for training/evaluation). State‑of‑the‑art in its class, with System 1, guard, multimodal variants and on‑device deployment. EUROPA, a 1 T‑parameter MoE, now training on the JUPITER supercomputer under the EuroHPC Frontier AI Grand Challenge. Domyn Agents, an open‑source, model‑agnostic agent harness with a safety and governance‑native design. In the talk we’ll cover pre/post‑training on public HPC, edge vs. frontier trade‑offs, and how an enterprise platform integrates these components for regulated sectors and acts as a flywheel for specialization.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4885,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Louis Powell",
    "title": "Director of AI Technologies",
    "company": "GSMA"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4955,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "API Standards Conference Welcome and OpenAPI in the Age of AI",
  "speakers": [
   {
    "name": "Erik Wilde",
    "title": "Chief Evangelist",
    "company": "WunderGraph"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "No abstract; OpenAPI's role when AI is the consumer.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4613,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "11:00",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Google Workshop",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Sponsor workshop, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4626,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "11:00",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Let the model make decisions. Let code move the data.",
  "speakers": [
   {
    "name": "Samuel Gaus",
    "title": "Principal Engineer and Team Lead",
    "company": "WunderGraph"
   },
   {
    "name": "Rom Kadria",
    "title": "Tech Lead",
    "company": "Monday.com"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality"
  ],
  "why": "Sponsor workshop; model decides, code moves data, a pattern for agent tool design.",
  "abstract": "Wundergraph Workshop",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4621,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "11:00",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Legitimate Requests, Illegitimate Intent: The Rise of Token-Based API Fraud",
  "speakers": [
   {
    "name": "Théo Triep-Capdeville",
    "title": "Presales Engineer",
    "company": "Thales"
   }
  ],
  "facets": [
   "operational_transparency",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity"
  ],
  "why": "Token-based API fraud and exposed credentials; security posture.",
  "abstract": "APIs are only as secure as the tokens protecting them. Explore emerging authentication risks, exposed credentials, and attack patterns, and learn how organizations can strengthen API security before attackers exploit them.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4604,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "11:00",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Let's Build, Secure, and Observe an AI Agent with an Agent Gateway",
  "speakers": [
   {
    "name": "Gbadebo Bello",
    "title": "Senior Developer Relations Advocate",
    "company": "Postman"
   }
  ],
  "facets": [
   "agent_readiness",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity",
   "rate_limit_signal"
  ],
  "why": "Routing an agent workflow through a gateway for credentials, tool access and observability.",
  "abstract": "AI agent workflows often begin as direct connections: the orchestration layer holds separate credentials, calls models and services independently, and manages tool access and reliability itself. In this hands-on workshop, you'll take an existing Banking, Fraud, and Support agent workflow and route it through a single governed layer using Fabric Gateway. You'll configure the gateway to connect the workflow's LLM, MCP servers, and APIs; centralize upstream credentials; shape and progressively disclose the tool surface; and apply gateway-level controls for routing, retries, and observability. Using the accompanying comparison app, you'll see the difference between the direct and gateway-mediated paths in real time. By the end, you'll understand how to move a working agent orchestration from scattered service connections to a unified, governed execution path. Prerequisites: Please come expecting to participate in a hands-on workshop where you can build with your peers. Laptops open, hands on keyboards welcome!",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4300,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "11:00",
  "end": "12:15",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Group work: What's Blocking API & Integration Success?",
  "speakers": [
   {
    "name": "Daniel Kocot",
    "title": "Principal API Consultant",
    "company": "adorsys"
   },
   {
    "name": "Claire Barrett",
    "title": "Founder",
    "company": "APIsFirst"
   },
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   },
   {
    "name": "Marjukka Niinioja",
    "title": "API Consultant & APIOps Community lead",
    "company": "Osaango Ltd"
   },
   {
    "name": "Miriam Greis",
    "title": "Principal Integration Architect.",
    "company": "adorsys GmbH + Co. KG"
   },
   {
    "name": "Miguel Quintero",
    "title": "Technical Trainer",
    "company": "Postman"
   },
   {
    "name": "Aiswarya Balasubramanian",
    "title": "API Architect",
    "company": "Anna University"
   }
  ],
  "facets": [
   "contract_governance",
   "discoverability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Group work on shadow APIs, governance and developer experience blockers.",
  "abstract": "Choose the APIOps station closest to your current challenge and work it through with peers facing similar problems. Topics are shaped by the room: platform adoption, shadow APIs, governance, executive buy-in, security, developer experience, or the problems participants bring with them. What happens Join one of the highest-priority station clinics Bring your own operating-model problem Diagnose what is actually blocking progress Outcome Clearer diagnosis Useful peer patterns Questions worth taking forward",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4921,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:00",
  "end": "12:25",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "Grep Is Not a Memory",
  "speakers": [
   {
    "name": "Lishuai Jing",
    "title": "co-founder, CAIO",
    "company": "memtrace.io"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; agent memory design, nothing a provider publishes.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4919,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "The problem isn't MCP. It's how we build MCP servers.",
  "speakers": [
   {
    "name": "Sébastien Charrier",
    "title": "CEO",
    "company": "Bump.sh"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "mcp_tool_quality"
  ],
  "why": "No abstract; building MCP servers well.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4820,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Remi Cadene",
    "title": "CEO",
    "company": "Uma"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4813,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Victor Schmidt",
    "title": "CTO",
    "company": "Entalpic"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4812,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Félix Raimundo",
    "title": "CEO",
    "company": "TychoBio.ai"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4722,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "GraphQL Variables Meet JSON Schema: The New Pitfall When AI Agents Call Your API",
  "speakers": [
   {
    "name": "Camille Lawrence",
    "title": "Senior Software Engineer",
    "company": "Apollo GraphQL"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "error_semantics"
  ],
  "why": "GraphQL variables and JSON Schema pitfalls when agents call an API; Arazzo workflows as portable contracts.",
  "abstract": "And as Arazzo is an open specification it makes those improvements portable and vendor-independent, turning enterprise workflow knowledge into durable, auditable infrastructure for the agentic era.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4795,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "Three common mistakes when taking an AI agent from prototype to production",
  "speakers": [
   {
    "name": "Matthieu Cordier",
    "title": "Applied AI Engineer",
    "company": "OpenAI"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Taking an agent from prototype to production; agent engineering, not a provider surface.",
  "abstract": "A demo shows that an AI agent can work once. How do you know it will work reliably on real tasks? How do you design a process to continually improve your system based on real usage? I’ll walk through three common mistakes teams make when building AI agents. You’ll leave with practical ways to simplify your AI agent system and make it more reliable, along with a sense of how to automate that improvement process over time.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4804,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "Inside Common Crawl: Open Web Data at Petabyte Scale",
  "speakers": [
   {
    "name": "Thom Vaughan",
    "title": "Principal Engineer",
    "company": "Common Crawl"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; open web crawl data at scale.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4788,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Raffi Krikorian",
    "title": "Mozilla",
    "company": "CTO"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4821,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "LeRobot: Making AI Robotics more accessible with end-to-end learning",
  "speakers": [
   {
    "name": "Steven Palma Morera",
    "title": "Tech Lead for LeRobot",
    "company": "Hugging Face"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; robotics learning, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4920,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "How to design an MCP for agents, avoid the pitfall of the API wrapper",
  "speakers": [
   {
    "name": "Thomas Payet",
    "title": "COO",
    "company": "Meilisearch"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "mcp_tool_quality",
   "operation_distinctiveness"
  ],
  "why": "No abstract; designing MCP for agents rather than wrapping the API one-to-one.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4796,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "They Said It Couldn't Be Done: Training SOTA LLMs On Fully Open Data",
  "speakers": [
   {
    "name": "Anastasia Stasenko",
    "title": "Co-Founder and CEO",
    "company": "pleias"
   }
  ],
  "facets": [
   "open_source",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; training LLMs on fully open data.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4723,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Securing Open Source CI/CD Pipelines: Lessons from Handling 2 Supply Chain Incidents",
  "speakers": [
   {
    "name": "Florence Njeri",
    "title": "Cybersecurity Engineer",
    "company": "AsyncAPI Initiative"
   },
   {
    "name": "Lukasz Gornicki",
    "title": "Open Source Maintainer",
    "company": "Brain Fart Services"
   }
  ],
  "facets": [
   "open_source",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "Supply chain incidents in open-source CI/CD; release and disclosure practice.",
  "abstract": "Open-source packages are now the primary attack surface used by threat actors. Rather than targeting enterprises directly, attackers are now compromising widely-used packages with a goal of reaching millions of CI/CD systems simultaneously through trusted releases. Because these open-source packages form the backbone of modern enterprise workflows, including cloud pipelines, CI/CD systems, and developer tooling, attackers are increasingly targeting open-source organizations to poison these packages and publish them to trusted registries such as npm and the VS Code Marketplace, compromising hundreds of downstream organizations simultaneously through trusted releases installed by their developers. Over the past year, AsyncAPI faced two supply chain attacks by TeamPCP and Miasma , compromising packages with up to 4M+ combined weekly downloads, such as: @asyncapi/specs (4M+ weekly) @asyncapi/generator (~130k weekly) Official VS Code extensions This session presents the tactics, techniques, and procedures (TTPs) we saw used in both incidents. We will trace how a single compromised release propagates into enterprise and public sector environments across Europe and France. Finally, attendees will gain actionable defense strategies and tools we have been using to tighten our security measures. We will showcase how we overhauled our release pipelines using: OIDC Trusted Publishing Automated CI/CD security scanning of every workflow change before it can release Hardened token isolation SHA-pinned GitHub Actions Least-privilege permission scopes on automation bots Attendees will leave with practical resources and tools to audit their own GitHub Actions, build pipelines, and IDE extension supply chains against active threat groups. The Technical Incidents: This section of the talk will provides first-hand incident response account from the AsyncAPI Initiative security team following two real-world software supply-chain compromises orchestrated by threat groups TeamPCP and Miasma. Exploitation of GitHub Actions pull_request_target workflows (pwn requests) to steal elevated repository secrets. Spam pull requests used to hide the real attack in the workflow logs, and a docs PR using an MDX file to run attacker code in our build pipeline. How the attackers released compromised packages with trusted credentials releasing malicious builds of @asyncapi/specs, @asyncapi/generator, and developer VS Code extensions directly to npm and the VS Code Marketplace, including one release through our own OIDC pipeline. Direct Impact of Supply Chain Attacks for Financial Sector Attendees in France and the EU: For banking, fintech, and public sector ecosystems in France and the EU operating under strict EU regulatory frameworks like DORA and NIS2, open-source packages and developer extensions represent an unmonitored third-party dependency vector. A poisoned upstream library imported into an internal financial pipeline or a developer’s IDE circumvents traditional perimeter controls. Actionable Takeaways: Attendees will leave with practical blueprints we implemented to remediate and harden our release infrastructure and further insights on how to protect their organizations from being compromised through supply chain attacks.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4805,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "Data, up for grabs or should be budgeted for?",
  "speakers": [
   {
    "name": "Leonora Vesterbacka",
    "title": "Founding Data Scientist",
    "company": "Redpine"
   }
  ],
  "facets": [
   "access_clarity",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; data as a budgeted cost, closest to finops.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4605,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "12:00",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Risk-Proof Your AI: The New API Governance Playbook",
  "speakers": [
   {
    "name": "Markus Müller",
    "title": "Global CTO - API Management",
    "company": "Boomi"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness",
   "accountability"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality"
  ],
  "why": "APIs as trusted agent tools, governance for enterprise readiness, measured for ROI.",
  "abstract": "Learn how to design APIs as trusted tools for AI agents, apply governance to make them enterprise-ready, and measure them for business impact and ROI.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4614,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "12:00",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Gravitee Workshop",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Sponsor workshop, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4622,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "12:00",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "API Security: Attack Lab Showdown",
  "speakers": [
   {
    "name": "Phil Ray",
    "title": "Enterprise API Security Specialist",
    "company": "Akamai"
   }
  ],
  "facets": [
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "Sponsor demo of API attack defense against agentic threats.",
  "abstract": "Explore API defense against agentic AI threats. Watch a live showdown of manual vs. automated attacks, see them in Akamai API Security, and learn strategies to protect your ecosystem.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4625,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "12:00",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Make Agentic AI Enterprise-Ready: Secure, Govern, and Observe Every Interaction",
  "speakers": [
   {
    "name": "Thierry Marchandise",
    "title": "Principal Pre-Sales Architect",
    "company": "Axway"
   },
   {
    "name": "Anuj Kumar",
    "title": "Principal Product Manager",
    "company": "Axway"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "auth_clarity"
  ],
  "why": "Securing, governing and observing agent, MCP and API interactions.",
  "abstract": "Connect agents, MCP, APIs, and enterprise systems through a controlled and repeatable architecture.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4301,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "12:15",
  "end": "13:00",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Group work: Choosing How to Tackle the Challenges",
  "speakers": [
   {
    "name": "Marjukka Niinioja",
    "title": "API Consultant & APIOps Community lead",
    "company": "Osaango Ltd"
   },
   {
    "name": "Claire Barrett",
    "title": "Founder",
    "company": "APIsFirst"
   },
   {
    "name": "Daniel Kocot",
    "title": "Principal API Consultant",
    "company": "adorsys"
   },
   {
    "name": "Miriam Greis",
    "title": "Principal Integration Architect.",
    "company": "adorsys GmbH + Co. KG"
   },
   {
    "name": "Aiswarya Balasubramanian",
    "title": "API Architect",
    "company": "Anna University"
   }
  ],
  "facets": [
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Group work picking one measurable change; operating model, no surface.",
  "abstract": "Do not redesign the whole operating model. Find one API, integration, event, product journey, or capability where a different way of working could demonstrate measurable improvement. Practitioner challenge Is it small enough to try? Important enough to matter? Can you tell whether it worked? What happens Identify one small enough proof capability or comparable experiment Decide what evidence would make it convincing Connect the experiment to adoption, controls, platform value, or business impact Outcome One candidate experiment An idea of what evidence would make it credible",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4942,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:30",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "AI model for Maps",
  "speakers": [
   {
    "name": "Charles-Pierre Astolfi",
    "title": "DSI",
    "company": "IGN"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; AI mapping model, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4814,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Geoffray Nejiar",
    "title": "CEO",
    "company": "The Forecasting company"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4918,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "MCP: Where Agents Meet Enterprise Governance",
  "speakers": [
   {
    "name": "José Carlos Chávez",
    "title": "Security Software Engineer",
    "company": "Okta"
   },
   {
    "name": "Anuj Kumar",
    "title": "Principal Product Manager",
    "company": "Axway"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "mcp_server"
  ],
  "why": "No abstract; MCP under enterprise governance.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4797,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "Swisscheese: the missing review process of your AI software factory",
  "speakers": [
   {
    "name": "Djalel Benbouzid",
    "title": "AI Governance, Phd",
    "company": "Independent AI Researcher"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Review process for AI coding factories; internal engineering, no provider surface.",
  "abstract": "AI software factories are everywhere right now. They are also the logical next step: models can now carry out long-running tasks, and agent harnesses have matured enough to run many of them in parallel. But models still make unpredictable mistakes, and harnesses are deliberately built to be as generic as possible. Scaling from a couple of Claude Code terminals to a real agent fleet multiplies both problems. Software engineering, and engineering in general, has dealt with unreliable components before. One of its lessons is the Swiss cheese model of risk management: no single layer of defence is perfect, so you stack several whose holes don't line up. In this talk, we apply that model to coding agents. We rethink parts of the harness around what current models do well and where they fail, and end up with a package that scales agent-based code review massively and uses agents' ability to supervise and correct each other. The result: Swiss cheese software.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4822,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Voice is the new OS",
  "speakers": [
   {
    "name": "Neil Zeghidour",
    "title": "CEO",
    "company": "Gradium"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; voice interfaces, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4724,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Make Your API Metadata AI-Ready and Automatically Discoverable by Agents",
  "speakers": [
   {
    "name": "Galina Tuova",
    "title": "Product Manager",
    "company": "SAP SE"
   }
  ],
  "facets": [
   "discoverability",
   "agent_readiness",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "well_known_catalog",
   "spec_presence"
  ],
  "why": "Open Resource Discovery: systems publish machine-readable metadata at the source for catalogs and agents.",
  "abstract": "Open Resource Discovery (ORD) addresses this with an open standard — governed by the Linux Foundation (NeoNephos Project) — for decentralized, self-describing systems. Instead of registering everything in a central catalog by hand, systems publish machine-readable metadata at the source, so catalogs, tools, and AI agents can crawl and refresh it automatically from a single entry point. ORD complements OpenAPI and AsyncAPI rather than replacing them, adding the high-level context and relationships that make APIs — along with events, data products, and AI resources such as agents, skills, and workflows — automatically discoverable, together with their contracts and documentation. ORD is also built for AI consumers: dedicated AI-facing metadata (such as hints for LLMs and agent orchestrators) and native support for agent protocols let AI orchestrators dynamically discover the capabilities available to them in a system landscape. With ORD, organizations can publish their metadata offerings, analyze their IT landscape in real time, and build platforms for software development, data management, and AI.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4789,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "AI for critical systems",
  "speakers": [
   {
    "name": "Juliette (Mayer) Mattioli",
    "title": "Thales Artificial Intelligence Fellow, VP. Algo/AI for Systems & Algo/AI Engineering Segment Leader",
    "company": "Thales"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; AI for critical systems, too vague to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4798,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "The first real Vibe coding Platform inside a corporate ? How TGV Agentic Team made it work, open to everybody and enabling software sovereignty",
  "speakers": [
   {
    "name": "Damien Wattez",
    "title": "Directeur IA et Innovation TGV",
    "company": "SNCF Voyageurs"
   },
   {
    "name": "Alexandre Cormeraie",
    "title": "Responsable IA & Innovation",
    "company": "SNCF Voyageurs"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; internal vibe-coding platform at a rail operator.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4922,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "Context Engineering: The Harness Decides What to Trust",
  "speakers": [
   {
    "name": "Ruben Marcus Luz Paschoarelli",
    "title": "Senior AI Fullstack Engineer",
    "company": "Independent"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; how agent harnesses decide what context to trust.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4877,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Beyond Validation: Reasoning About What a Schema Means",
  "speakers": [
   {
    "name": "Jason Desrosiers",
    "title": "JSON Schema Specification and Tooling Architect",
    "company": "Hyperjump Software"
   }
  ],
  "facets": [
   "contract_quality",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [],
  "why": "Reasoning about what JSON Schema means: contradictions, satisfiability, beyond validation.",
  "abstract": "Most JSON Schema implementations evaluate keywords. A validator checks a value against \"\"minimum\"\" or \"\"properties\"\" and reports whether it passed. That's enough for validation, but other questions need more. Does this schema allow anything at all? Does this \"\"minimum\"\" contradict that \"\"maximum\"\"? Is this required property one that \"\"additionalProperties: false\"\" forbids? Answering those means reasoning about what a schema means, not just evaluating it. Version 2 of @hyperjump/json-schema is built around that idea. Every keyword, in every dialect from draft-04 to JSON Schema v1, lowers to a kernel: a small language of operations with a normal form. The validator runs the kernel, the optimizer rewrites it, and the output formats explain it. The linter's 48 rules read it to answer questions like the ones above. None of them keeps its own facts about keywords. What a keyword does comes from what it lowers to, so a rule written once works in every dialect. That includes custom keywords. A custom keyword is defined by what it lowers to, so the linter and optimizer understand it like a built-in one, with no extra code. Understanding a schema also makes it fast. On the jsonschema-benchmark corpus, v2 validates 14 times faster than v1 and about 4 times faster than Ajv. It does that without generating code: no \"\"eval\"\" or \"\"new Function\"\", so it runs under a strict Content Security Policy. This talk shows what lowering looks like and how it makes dialects comparable. It covers how the optimizer gets speed without code generation, and how lint rules work, down to the kernel expressions they read. Attendees will leave with a way of thinking about schema semantics they can use in their own implementations and tools.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4807,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "Beyond General-Purpose Models: Building AI That Delivers Real-World Value",
  "speakers": [
   {
    "name": "Olivier Debeugny",
    "title": "CEO",
    "company": "Ovhai LLM"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; domain AI models, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4823,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "AI Edge for the AI Age: Streaming Data for Robots and Drones",
  "speakers": [
   {
    "name": "Jean-Baptiste Kempf",
    "title": "President, VideoLAN / CEO",
    "company": "Kyber / Tech Fellow Scaleway"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "No abstract; streaming data for robots and drones at the edge.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4790,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Safe to leave running: the engineering behind autonomous agents",
  "speakers": [
   {
    "name": "Paul-Emile Brotons",
    "title": "Member of Technical Staff",
    "company": "Anthropic"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "reversibility_documented",
   "dry_run_mode",
   "idempotency"
  ],
  "why": "No abstract; engineering agents safe to leave running unattended.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4627,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Microsoft Workshop",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Sponsor workshop, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4615,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Scale your AI agents, Keep control: A Live Demo from foundation to production",
  "speakers": [
   {
    "name": "Isabelle Mauny",
    "title": "Field CTO",
    "company": "Choreo Labs, formerly WSO2"
   },
   {
    "name": "Mohammed Riahi",
    "title": "Solutions Architect",
    "company": "Choreo Labs, formerly WSO2"
   }
  ],
  "facets": [
   "agent_readiness",
   "access_clarity",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity",
   "rate_limit_signal"
  ],
  "why": "Governed fleet of agents with security, compliance and cost control over models and APIs.",
  "abstract": "Every team is now building agents, and each one connects to models, tools and APIs on its own. Without a common foundation, security gaps, compliance exposure and runaway costs grow faster than the agents themselves. This live demo takes you from the first controlled LLM connection to a governed fleet of agents running in production. Step by step, we lay the foundation and show how to add what an enterprise needs to scale: ● Security: agent identity, access control and guardrails. ● Compliance: audit trails and policy enforcement. ● Governance: visibility on which agents exist, what they call and who owns them. ● FinOps: usage tracking, quotas and cost control per team, agent and model. You will leave with a practical blueprint to move agents from experiments to production, with control built in by design rather than added after the first incident.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4623,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Deploying and Managing Layer7 within a Kubernetes environment",
  "speakers": [
   {
    "name": "Stephane Murer",
    "title": "Technical Leader",
    "company": "Malakoff Humanis"
   },
   {
    "name": "Gary Vermeulen",
    "title": "Head of Layer7 Architecture",
    "company": "Broadcom"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Running a gateway on Kubernetes; internal deployment, no published surface.",
  "abstract": "Stéphane Murer, Technical Leader at Malakoff will discuss key insights and practical aspects of deploying and managing Layer7 within their Kubernetes environment.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4606,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Let's Launch a Rocket: AI-Powered API & MCP Bootcamp",
  "speakers": [
   {
    "name": "Anthony Viard",
    "title": "Senior Developer Advocate",
    "company": "Postman"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "spec_presence"
  ],
  "why": "AI tooling from spec to collection to code, plus MCP.",
  "abstract": "With the rise of Large Language Models, AI copilots, and autonomous agents, developers can now design, test, and iterate on APIs with greater speed and granularity than ever before. Modern AI tooling makes it possible to move seamlessly from spec to collection and from collection back to code, accelerating every stage of the API lifecycle. This 50-minute, hands-on bootcamp is designed to empower developers who want to build, test, and automate APIs and Model Context Protocol (MCP) servers using AI-powered workflows. You'll work in an environment where AI is always available, leveraging Postman Agent Mode to prototype, validate, and automate real, working API systems. You'll move quickly from concept to execution. In this session, you will: Prototype, test, and automate API workflows using Postman Agent Mode Generate environments, scripts, and documentation by describing what you need, not writing it Visualize and validate live API responses as rich, interactive dashboards Build, test, and validate an end-to-end integration test suite from a working API Understand how AI agents interact with APIs and MCP-enabled services in practice By the end of the bootcamp, you'll have hands-on experience with a complete, end-to-end workflow for building and validating AI-ready APIs using tools and techniques you can immediately apply in your own projects. Prerequisites: Please come expecting to participate in a hands-on workshop where you can build with your peers. Laptops open, hands on keyboards welcome! Basic knowledge of APIs (HTTP, REST, or experience working with an API) A Postman account (sign up free) Postman Desktop App recommended but not mandatory Familiarity with API design or testing workflows is helpful but not required No prior AI or LLM experience required",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4619,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Upsun Workshop",
  "speakers": [
   {
    "name": "Guillaume Moigneu",
    "title": "Field CTO",
    "company": "Upsun"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Sponsor workshop, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4302,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "14:15",
  "end": "16:00",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Group work: From API Challenges to Action Plans",
  "speakers": [
   {
    "name": "Claire Barrett",
    "title": "Founder",
    "company": "APIsFirst"
   },
   {
    "name": "Daniel Kocot",
    "title": "Principal API Consultant",
    "company": "adorsys"
   },
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   },
   {
    "name": "Marjukka Niinioja",
    "title": "API Consultant & APIOps Community lead",
    "company": "Osaango Ltd"
   },
   {
    "name": "Miriam Greis",
    "title": "Principal Integration Architect.",
    "company": "adorsys GmbH + Co. KG"
   },
   {
    "name": "Miguel Quintero",
    "title": "Technical Trainer",
    "company": "Postman"
   },
   {
    "name": "Aiswarya Balasubramanian",
    "title": "API Architect",
    "company": "Anna University"
   }
  ],
  "facets": [
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Ownership, actions and evidence for a 90-day move; operating model, no surface.",
  "abstract": "Turn the morning's diagnosis into ownership, actions, evidence, and a credible 90-day move. Do not design the transformation. Define the first credible move and what needs to happen in the first 30 days. What happens Work through ownership, stakeholders, incentives, controls, and blockers Define the first 30 days Clarify the evidence needed over a rough 90-day path Outcome A rough 90-day path First 30-day actions Evidence and risk notes",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4725,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "One to many without breaking anyone: evolving a GraphQL schema with many consumers",
  "speakers": [
   {
    "name": "Gabriela Colombo Alves Longo",
    "title": "Software Engineer II",
    "company": "Swapcard"
   }
  ],
  "facets": [
   "operational_transparency",
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "error_semantics"
  ],
  "why": "Evolving a schema without breaking consumers; Arazzo workflows with success criteria and failure handling.",
  "abstract": "Arazzo changes the unit of exchange. By composing API operations into business-defined workflows with typed inputs, success criteria and structured failure handling, Arazzo lets organisations expose what agents actually need: goal-oriented capabilities, not a catalogue of primitives. Those workflows become first-class MCP tools — derived directly from the Arazzo document, with no hard-coded tool definitions to write or lifecycle to manage. Descriptions, typed schemas and examples are generated from the workflow itself, reducing search-and-execute overhead and guiding agents toward completion rather than exploration.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4808,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Nicolas Flores Herr",
    "title": "Team Lead Foundation Models & Gen AI Systems / Standortleiter",
    "company": "Fraunhofer IAIS Dresden"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4791,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Panel",
  "speakers": [
   {
    "name": "Roxanne Varza",
    "title": "Director, Station F",
    "company": "Station F"
   },
   {
    "name": "Romain Dillet",
    "title": "VC",
    "company": "Drysdale Ventures"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; untitled panel, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4787,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Model Factory",
  "speakers": [
   {
    "name": "Pierre-Yves Ritschard",
    "title": "CTO, Poolside",
    "company": "NVIDIA"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; model factory, no API relevance in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4815,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Laurie Soffiati",
    "title": "CEO",
    "company": "Nabla"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4799,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "Can agents learn from their mistakes?",
  "speakers": [
   {
    "name": "Jules Belveze ",
    "title": "Core Team Researcher",
    "company": "H Company"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Agents learning from failed runs; agent research, no provider surface.",
  "abstract": "Agents make mistakes. The frustrating part is watching them make the same ones again. What would it take for an agent to learn from a failed run? Agents can use past runs to revise their instructions, tools, or memory without retraining the model. Recent research explores different ways to turn that experience into useful changes. The tricky part is checking whether a change actually helps. It might fix one case and break another, or improve a score while leaving the underlying problem untouched. Learning from mistakes means testing proposed fixes and keeping evidence of what worked, what failed, and why.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4923,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "Your Agent Is Authorized. Your Data Isn't.",
  "speakers": [
   {
    "name": "Julian Kusenberg",
    "title": "Senior Consultant / Microsoft MVP",
    "company": "SoftwareOne"
   }
  ],
  "facets": [
   "agent_readiness",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "delegated_identity",
   "consent_identity"
  ],
  "why": "No abstract; authorized agents still need data-level authorization.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4816,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Anna Hlédiková",
    "title": "CEO",
    "company": "SurgeryAI"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA, nothing to map.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4915,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Your APIs Are Ready. But Are They AI-Ready? From API Sprawl to AI-Ready Governance - A Live Journey with Solis Hub",
  "speakers": [
   {
    "name": "Iliass Ayaou",
    "title": "Chief Research Officer",
    "company": "GroupBees"
   },
   {
    "name": "Youssef Guenoun",
    "title": "Co-Founder and CTO",
    "company": "GroupBees"
   }
  ],
  "facets": [
   "contract_governance",
   "discoverability",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "Sponsor workshop; API sprawl to governed, discoverable, AI-ready APIs.",
  "abstract": "APIs are everywhere. Governance isn't. As AI agents take over, who's in control? Join our live Solis Hub workshop to turn API chaos into a secure, governed, discoverable and AI-ready ecosystem.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4726,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Beyond the Broker: What Event Gateways Are Really Trying to Solve",
  "speakers": [
   {
    "name": "Daniel Kocot",
    "title": "Principal API Consultant",
    "company": "adorsys"
   }
  ],
  "facets": [
   "contract_governance",
   "contract_quality",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "Event gateways: ownership, compatibility over time and policy for events.",
  "abstract": "Event gateways are becoming a popular answer to a real problem in event-driven architecture: the lack of a clear control plane. But they should not be understood as API gateways for events. Events have different failure modes. Ownership is less visible. Compatibility changes over time. Policies are harder to enforce without coupling producers and consumers again. Observability depends on more than traffic flowing through a single point. This talk examines where event gateways are useful, where brokers, schema registries, catalogs, and portals already cover part of the problem, and where the gateway metaphor starts to break down. The focus is not another infrastructure layer. The focus is which governance capabilities event-driven systems actually need, and how AsyncAPI can help describe them.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4951,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Build Your Governed Context Layer",
  "speakers": [
   {
    "name": "Zaak Chalal",
    "title": "CEO",
    "company": "Mobioos"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Connect specs as a governed context layer so agents stop guessing.",
  "abstract": "Build a governed context layer in 50 minutes. Connect your specs, apply them to code in your IDE and see AI agents stop guessing. Bring a laptop with VS Code, Cursor or Kiro.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4824,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Every Call Is Opening Night for AI Agents",
  "speakers": [
   {
    "name": "Italo Vietro",
    "title": "VP of Engineering",
    "company": "Parloa"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; agents hitting APIs live on every call.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4792,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Gamification of Large Language Models",
  "speakers": [
   {
    "name": "Michal Valko",
    "title": "Principal Engineer, LLama Team, GenAI",
    "company": "Meta"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; LLM gamification, no API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4809,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "Developing fully open European language models with a French twist",
  "speakers": [
   {
    "name": "Julie Hunter",
    "title": "Research Engineer",
    "company": "Linagora R&D Team"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; open European language models, no API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4956,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "Running a Frontier AI Research Lab with 5 Researchers… and Agents",
  "speakers": [
   {
    "name": "Jean Du Terrail",
    "title": "Head of AI Research",
    "company": "Living Models"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Running an AI research lab with agents; no API surface.",
  "abstract": "It has become abundantly clear that the impressive capabilities of AI agents can be harnessed to streamline research in most scientific domains, including AI research itself, and the words \"\"recursive self-improvement (RSI)\"\" are on everyone's lips. Yet frontier labs share surprisingly few details about how to do this at scale, notably, very practically, how to organize research around this new multi-agents technology. This talk will take you inside our biological AI model factory at Living Models where 5 AI researchers collaborate with multiple AI agents to create, train and evaluate state-of-the-art foundation models on DNA. In this talk you'll learn, among other things, why biology is so different from other domains for AI research, why it won't be \"\"RLVR-ed\"\" (learned through reinforcement learning with verifiable rewards) as easily as code, and what we do about it at Living Models.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4924,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "The Context Supply Chain: Engineering Reliable Context for AI Agents",
  "speakers": [
   {
    "name": "Daniel Jarjoura",
    "title": "Founder",
    "company": "Rubbr"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "operation_distinctiveness",
   "mcp_tool_quality"
  ],
  "why": "No abstract; engineering the context agents consume.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4800,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "API & marketing for agents",
  "speakers": [
   {
    "name": "Elie Chevignard",
    "title": "Clarity for Deep Tech | Product Marketing | PhD",
    "company": "CEA"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; marketing APIs to agents is machine-readable findability.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4607,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "15:00",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "How Kong Konnect keeps your AI spend under control",
  "speakers": [
   {
    "name": "Jerome Guillaume",
    "title": "Staff Solutions Engineer",
    "company": "Kong"
   }
  ],
  "facets": [
   "access_clarity",
   "operational_transparency",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "rate_limit_signal"
  ],
  "why": "Sponsor demo, but token tracking, limits and cost control is finops.",
  "abstract": "AI bills creep up unseen. Jerome Guillaume shows how Kong Konnect tracks token usage, sets limits, and routes requests to the right model, so you control AI spend without slowing developers down.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4616,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "15:00",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Federated API Intelligence in Action: A Hands-On Demo with Treblle",
  "speakers": [
   {
    "name": "Vedran Cindrić",
    "title": "Founder & CEO",
    "company": "Treblle"
   }
  ],
  "facets": [
   "operational_transparency",
   "discoverability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Sponsor demo of API visibility across an estate.",
  "abstract": "Every company is becoming an API company, and AI is speeding that up. In this interactive demo, Treblle's founder and CEO shows how Treblle gives your teams full visibility into the APIs behind your digital transformation. You'll see how it cuts manual work, reduces risk, and gets you ready for AI, so you can move faster.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4925,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "ContextCon and MCP Summit"
  ],
  "room": [
   "Dragonnier 3"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Quentin de Quelen",
    "title": "CEO & Co-founder",
    "company": "Meilisearch"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4878,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Arazzo: Turning Business Workflows into Agent Capabilities via MCP and Beyond",
  "speakers": [
   {
    "name": "Frank Kilcommins",
    "title": "Head of Enterprise Architecture",
    "company": "Jentic"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "mcp_tool_quality"
  ],
  "why": "Arazzo workflows as agent capabilities via MCP.",
  "abstract": "Giving agents raw API operations is like handing someone a parts list instead of a finished product. Individually the operations are valid but assembled in the wrong order, with the wrong inputs, they produce the wrong outcomes. This is wasteful, unpredictable, and incurs a real cost on both sides of the interaction.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4793,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Steeve Morin",
    "title": "ZML",
    "company": "CEO"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4801,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Serena Villata",
    "title": "Head of the MARIANNE research team",
    "company": "INRIA"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4817,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "Panel",
  "speakers": [
   {
    "name": "Axel Colin De Verdière",
    "title": "Engineering Director",
    "company": "Doctolib"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; untitled panel.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4802,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "Panel : Vers des IA de recommandation démocratiques – Le projet Tournesol",
  "speakers": [
   {
    "name": "Jean Lou Fourquet",
    "title": "Co-Founder",
    "company": "Lyfe Catalyst"
   },
   {
    "name": "Lê Nguyên Hoang",
    "title": "Co-Founder & CEO",
    "company": "Calicarpa"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Democratic recommendation AI; no API surface.",
  "abstract": "Chaque jour, des milliards de personnes voient leur vision du monde façonnée par des algorithmes opaques qui décident de ce qu’elles voient en ligne. Ces intelligences artificielles de recommandation — moteurs invisibles de plateformes comme YouTube, TikTok ou Instagram — influencent profondément le débat public, l’accès à l’information et même la stabilité démocratique. Le projet Tournesol est né de trois constats fondamentaux : 1) Les systèmes de recommandation, tels qu’ils existent aujourd’hui, ont tendance à fragiliser nos démocraties. 2) Ces systèmes doivent être véritablement régulés et les plateformes qui les développent redevables. 3) Construire un espace numérique plus démocratique est un défi existentiel pour nos sociétés. Tournesol développe des systèmes de recommandation alternatifs et plus démocratiques, fondés sur les jugements et les intérêts collectifs des citoyens. Le projet recueille et agrège des évaluations humaines (volitions) — des jugements portés par de vraies personnes sur la qualité, la fiabilité et la valeur sociétale des vidéos YouTube, en les comparant entre elles. Entièrement open source, publiquement auditable, et guidé par des valeurs collectives plutôt que par des incitations commerciales, la plateforme Tournesol repose sur quatre piliers essentiels d'un numérique démocratique : 1) Transparence – Les systèmes de recommandation doivent être ouverts, vérifiables et compréhensibles par le plus grand nombre. Les algorithmes et les données de Tournesol sont publiquement accessiblesc, permettant aux citoyens et aux chercheurs de comprendre comment les évalutations sont faites. 2) Résilience face aux attaques – Les plateformes et leurs IA de recommandation doivent être protégées contre les manipulations massives menées par attaques concertées d'un très grand nombre de faux comptes. Tournesol intègre des mécanismes de détection et de résistance à ces menaces, garantissant une participation authentique. 3) Basé sur des avis réfléchis – Les recommandations devraient refléter les choix délibérés des utilisateurs plutôt que leurs réactions impulsives (clics, balayages, etc.). Tournesol privilégie l’évaluation volontaire et réfléchie aux signaux de préférence instantanés, favorisant une culture de la réflexion plutôt que de l’addiction. 4) Favorable à l'émergence d'un socle commun informationnel – Les espaces numériques devraient favoriser une compréhension commune du monde, plutôt que de fragmenter les utilisateurs dans des bulles hyper-personnalisées. Tournesol encourage l’exposition collective à des contenus de qualité, contribuant à reconstruire un socle de connaissances partagé. Au cours de la présentation de tournesol.app et de ses extensions, les participants découvriront : * Ce à quoi pourrait ressembler une gouvernance collaborative de la recommandation par l’IA ; * Comment Tournesol peut transformer leur expérience de YouTube ; * Comment le projet applique les quatre piliers d’un espace numérique démocratique, à l’inverse des grandes plateformes comme TikTok ; * Et quelles sont les prochaines étapes du projet pour renforcer encore son alignement avec ces principes. En définitive, Tournesol invite chacun à repenser la manière dont les algorithmes façonnent notre monde numérique partagé — non pas en simples consommateurs, mais en citoyens actifs.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4954,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Yasser Jadidi",
    "title": "Co-CRO",
    "company": "AlephAlpha"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4818,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "Data science in the age of AI",
  "speakers": [
   {
    "name": "Gaël Varoquaux",
    "title": "Co-founder",
    "company": "Probabl"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Data science with AI; no API surface.",
  "abstract": "LLMs have transformed how we process information, and Agentic AI is transforming coding. But data-science needs more than coding because it is much harder to verify. How can today's AI help us build reliable data-science models faster? I'll detail the challenges that a data-science agent must overcome, and how the emerging space of tabular AI provides answers combining new and old tools, including an update on scikit-learn.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4303,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Open Discussion: The Toughest API Questions We Still Face",
  "speakers": [
   {
    "name": "Claire Barrett",
    "title": "Founder",
    "company": "APIsFirst"
   },
   {
    "name": "Daniel Kocot",
    "title": "Principal API Consultant",
    "company": "adorsys"
   },
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   },
   {
    "name": "Marjukka Niinioja",
    "title": "API Consultant & APIOps Community lead",
    "company": "Osaango Ltd"
   },
   {
    "name": "Miriam Greis",
    "title": "Principal Integration Architect.",
    "company": "adorsys GmbH + Co. KG"
   },
   {
    "name": "Miguel Quintero",
    "title": "Technical Trainer",
    "company": "Postman"
   },
   {
    "name": "Aiswarya Balasubramanian",
    "title": "API Architect",
    "company": "Anna University"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Open discussion format with no stated topic.",
  "abstract": "Bring the hardest unresolved question from the station clinics back into the room. Not a panel Practitioners and participants rotate into the conversation. No panel. What happens Rotate into the discussion Challenge the assumptions that survived the clinics Debate the question the room decides deserves the final conversation Outcome New perspectives Assumptions challenged Open questions for the community",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4727,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Not Yet Valid: Four Questions Your Editor Asks a JSON Schema",
  "speakers": [
   {
    "name": "Bart Dubbeldam",
    "title": "Founding Engineer",
    "company": "Cached Intelligence"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "How editors validate half-typed JSON Schema documents.",
  "abstract": "A schema answers the question: is this finished document valid? However, an editor often sees an unfinished document. While someone types, the file is not yet valid, and the editor still has to answer four questions: Where am I in a half-typed, broken document? What could go here? What actually went wrong? What does this part of the schema mean? This talk works through each question using the JSON Schema support behind hover, completion, go-to-definition and diagnostics in KSON, an open-source configuration language. Some examples: an if that depends on a field you haven't typed yet. A single typo in a oneOf that gives you an error for every branch, even though const and enum could tell the editor which branch you meant. A property with three descriptions, and no rule for which one hover shows. Each section ends with a rule schema authors can apply today.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4810,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3/4"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Monsieur Moyen",
    "title": "Movie Director",
    "company": "Le Prompt Club"
   },
   {
    "name": "Stéphane Tranquillin",
    "title": "Movie Director",
    "company": "Le Prompt Club"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4826,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "16:30",
  "end": "17:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Panel: Vulgariser les enjeux de l’IA, comment informer ?",
  "speakers": [
   {
    "name": "Axelle Arquié",
    "title": "Economist",
    "company": "Observatory of Threatened and Emerging Jobs"
   },
   {
    "name": "Thibaut Giraud",
    "title": "Youtuber and Philosopher",
    "company": "Monsieur Phi"
   },
   {
    "name": "Jérémy Perret",
    "title": "Member",
    "company": "Coexiscience"
   },
   {
    "name": "Shaïman Thürler ",
    "title": "Fondateur et directeur",
    "company": "Le Futurologue"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; panel on explaining AI to the public.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4608,
  "day": "2026-12-01",
  "type": "Workshop",
  "start": "16:30",
  "end": "17:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Legal Waiting Tax: what your teams lose while waiting for legal",
  "speakers": [
   {
    "name": "Alex Hayem",
    "title": "CEO / Founder",
    "company": "MIRMI AI"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Legal delays and AI; no API surface.",
  "abstract": "Contracts, projects and decisions face legal delays every day. What does that really cost? Using a real case, measure your bottlenecks and see how AI can finally help.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4803,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "Guerillia: how can we take back power over artificial intelligence?",
  "speakers": [
   {
    "name": "Christophe Payet",
    "title": "Lab IA artists",
    "company": "Arte"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Taking back power over AI from Big Tech; no API surface.",
  "abstract": "Artificial intelligence is everywhere. Are we doomed to depend on Big Tech? Are there other ways to make these tools our own? Led by Christophe Payet, journalist and essayist, co-creator of PhantasIA on Arte and author of GuerillIA. Mener la bataille culturelle de l’intelligence artificielle (Éditions du Seuil), this talk goes beyond critique to ask what truly reclaiming AI could mean. How can we adopt other tools, uses and imaginaries? Against the myth of “superintelligence,” artists, activists and hackers are already diverting these technologies and inventing other ways to conceive and use them. Caught between blissful technosolutionism and a growing, wholesale rejection, this is a critical yet enthusiastic invitation to make AI a genuine tool for emancipation.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4879,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "oasdiff vs Claude: A Case Study in LLMs vs Dedicated Tools",
  "speakers": [
   {
    "name": "Reuven Harrison",
    "title": "CTO",
    "company": "oasdiff"
   }
  ],
  "facets": [
   "contract_governance",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; oasdiff is OpenAPI breaking-change detection.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4794,
  "day": "2026-12-01",
  "type": "Panel discussion",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Panel: l'IA au service de l’humain",
  "speakers": [
   {
    "name": "Stéphie Herlin",
    "title": "Co-Founder",
    "company": "KORA"
   },
   {
    "name": "Etienne Grass",
    "title": "Chief of AI",
    "company": "Cap Gemini"
   },
   {
    "name": "Ludovic Peran",
    "title": "Technical Strategy",
    "company": "OpenAI"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; panel on AI serving people.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4819,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 5"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Hector Gonzales",
    "title": "Co-Founder and CEO",
    "company": "SpiNNcloud"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4304,
  "day": "2026-12-01",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "APIOps Conference Paris 2026"
  ],
  "room": [
   "Davidia 1/2"
  ],
  "title": "Key Takeaways & Next Steps",
  "speakers": [
   {
    "name": "Marjukka Niinioja",
    "title": "API Consultant & APIOps Community lead",
    "company": "Osaango Ltd"
   },
   {
    "name": "Aiswarya Balasubramanian",
    "title": "API Architect",
    "company": "Anna University"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Workshop wrap-up, no API topic of its own.",
  "abstract": "What did we learn, what patterns emerged, and what will you do next? What happens Create the Paris APIOps Operating Notes Capture patterns and unresolved questions from the day Choose one station, one decision, and one next step Outcome Shared synthesis Personal next action Community learning that can continue after Paris",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4891,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "09:00",
  "end": "09:15",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Mathematics in the Age of AI",
  "speakers": [
   {
    "name": "Hugo  Duminil-Copin",
    "title": "Mathematicien",
    "company": "IHES"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Mathematics and AI; no API surface.",
  "abstract": "Generative artificial intelligence is transforming an ever-growing range of professions. Recently, it has succeeded in solving some of the most challenging mathematical conjectures. From software developers to artists, mathematicians and writers, no field of intellectual creativity seems beyond its reach. As AI becomes increasingly capable of automating complex tasks, it sometimes makes us lose sight of why we put so much effort into carrying them out in the first place. This talk will invite us to reflect on the role of human beings in education, research and innovation. What do we wish to preserve of our intelligence, curiosity and freedom in an age when artificial intelligence is pushing the boundaries of what is possible?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4628,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "09:00",
  "end": "10:00",
  "track": [
   "Apidays",
   "Women in API"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "GET /Connecting breakfast – Women in APIs",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Networking breakfast.",
  "abstract": "When you're facing a difficult decision, navigating a career change or trying to take your next step, who do you turn to? We often hear about the importance of finding a mentor, working with a coach or building a strong professional network. But knowing what kind of support you need, where to find it and how to ask can be unclear. Join us for an interactive breakfast exploring how we find guidance and support throughout our careers. Through facilitated roundtable conversations, we'll share experiences, discuss what holds us back from asking for help and explore how different relationships can help us move forward. Whatever stage you are at now, whether you are a mentor, mentee or unsure whether you need a support. Come with your experiences, questions and curiosity. Leave with fresh perspectives, practical ideas for finding the support you need and new connections who might just become part of your own support network. Everyone attending FOST Paris is welcome, regardless of gender.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4567,
  "day": "2026-12-02",
  "type": "Opening remarks",
  "start": "09:15",
  "end": "09:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Opening remarks · Bringing Tech Home: Build Less, Build Better",
  "speakers": [
   {
    "name": "Gael Duez",
    "title": "Founder & Host",
    "company": "Green IO"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Opening remarks, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4312,
  "day": "2026-12-02",
  "type": "Opening remarks",
  "start": "09:15",
  "end": "09:30",
  "track": [
   "Apidays",
   "GenerationAI"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Welcome and Opening Remarks",
  "speakers": [
   {
    "name": "Mehdi Medjaoui",
    "title": "Author of Continuous API Management",
    "company": "Founder and Chairman of Apidays Conferences"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Opening remarks, no abstract.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4884,
  "day": "2026-12-02",
  "type": "Keynote",
  "start": "09:30",
  "end": "09:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "What the history of AI tells us about its future",
  "speakers": [
   {
    "name": "Alexandre Gefen",
    "title": "AI Research",
    "company": "CNRS"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "History of AI; no API surface.",
  "abstract": "Today’s AI inherits two things from its past: a set of myths and a set of unsolved scientific questions. This 20-minute talk first walks through seven figures that predate the computer and still shape products, benchmarks and fears: the speaking statue of Amun (religion), Vaucanson’s duck (rivalling nature), the chess automaton (play), the Jaquet-Droz writer (art), Pygmalion (love), Talos (war) and the Golem (fear). It then follows six periods, from the calculating machines of the 17th century to the foundation models of 2018-2025, through cybernetics, the Dartmouth moment, the first winter, expert systems and the return of neural networks. Each stage ends with a short “looking ahead” box asking what it may suggest for the systems being built now, and the talk closes on five hypotheses rather than predictions. One more thing: “vintage” language models trained only on texts of the past, and FLAUBOT, a Flaubert chatbot, serve as a demonstration that tools built from the past can be used to test our ideas about the future. The talk draws on Histoire culturelle de l’IA (CNRS Éditions, 2026).",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4869,
  "day": "2026-12-02",
  "type": "Keynote",
  "start": "09:30",
  "end": "10:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "AI and planetary boundaries",
  "speakers": [
   {
    "name": "Philippe Bihouix",
    "title": "Managing Director",
    "company": "AREP"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI and planetary boundaries; nothing in the rubric reads it.",
  "abstract": "Techno-optimist promises are more pervasive than ever. If we are to believe the \"prophets\" of Silicon Valley, artificial intelligence will \"fix\" the climate and lead to a future of phenomenal abundance and wealth for all – driven by autonomous robots, neural chips and the conquest of space. Is this truly realistic – or even sensible –, given planetary boundaries, environmental degradation and widespread pollution, fast-growing demands for resources and energy, and geopolitical tensions? What if our energy transition and innovation efforts were to focus on different paths – ones that are maybe more \"restrained\" but far more resilient?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4658,
  "day": "2026-12-02",
  "type": "Keynote",
  "start": "10:00",
  "end": "10:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Thinking With Machines",
  "speakers": [
   {
    "name": "Mike Amundsen",
    "title": "Author of \"Design and Build Great APIs\", API Strategist & Advisor",
    "company": "amundsen.com, Inc."
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Thinking with machines; no API surface.",
  "abstract": "What is the future of human thinking in an AI-native world? As intelligent machines become part of how we work, design, and make decisions, how can we use them to expand human capability and creativity? Thinking with Machines revisits a longstanding idea in computing: machines should help people reason, explore, and decide, rather than simply automate outcomes. Building on the work of Vannevar Bush, J. C. R. Licklider, and Douglas Engelbart, this keynote explores what it means to design AI-native systems that treat machines as active partners in human reasoning. Through engaging stories and practical examples, the talk introduces patterns for AI-guided exploration, collaborative problem-solving, and decision support. It shows how thoughtful interaction design can extend human judgment and strengthen creative thinking while preserving human agency. The result is a practical vision for AI-native computing grounded in decades of research on human-machine collaboration and focused on making modern AI useful in complex, real-world work.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4572,
  "day": "2026-12-02",
  "type": "Keynote",
  "start": "10:00",
  "end": "10:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Why women need a seat at the table in the AI transition",
  "speakers": [
   {
    "name": "Anna Lerner",
    "title": "CEO",
    "company": "Climate Collective"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Women in the AI transition; no API surface.",
  "abstract": "Women use AI differently, have different approaches/tolerance to AI risks, face bigger risk for work replacement and are less frequently hired in AI companies (technical and leadership positions). We can stand by and watch - or decide to do something about it. Climate Collective is running women only cohorts for AI fluency, that's one path, but we need many more tactics and initiatives to change the status quo.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4659,
  "day": "2026-12-02",
  "type": "Keynote",
  "start": "10:30",
  "end": "10:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Amphitheater"
  ],
  "title": "Your API platform was built for humans. Nobody told the agents",
  "speakers": [
   {
    "name": "Harry Ault",
    "title": "CEO",
    "company": "Choreo Labs, formerly WSO2"
   }
  ],
  "facets": [
   "agent_readiness",
   "access_clarity",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "delegated_identity",
   "auth_clarity"
  ],
  "why": "API platforms built for humans now exposing MCPs to agents.",
  "abstract": "Every product team wants its MCP exposed to agents. Every BU wants access to SaaS MCPs. They all want it this quarter. The problem isn’t that your API team needs to become a platform team. You made that shift years ago. The problem is that your platform assumes consumers who read docs, onboard once, authenticate as an app, and call at human speed. Agents break every one of those assumptions, and adding people won’t close the gap. This keynote covers what has to move from manual process into the platform itself: MCP servers generated from existing APIs, governance in the CI pipeline, an identity for every agent, and policy enforced at the gateway. It draws on how Choreo Labs, formerly WSO2, customers in 90+ countries run 60 trillion transactions a year without a review meeting for each one.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4545,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "10:35",
  "end": "11:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Comment les entreprises de la Tech et leurs clients intègrent-ils la gouvernance de l’IA dans leurs pratiques ?",
  "speakers": [
   {
    "name": "Emmanuelle Olivié-Paul",
    "title": "Présidente-Fondatrice",
    "company": "AdVaes"
   }
  ],
  "facets": [
   "accountability",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "How companies build AI governance into practice.",
  "abstract": "L’évolution rapide et sans précédent de l’utilisation de l’IA représente un changement de paradigme que nul d’entre nous ne peut plus ignorer. Elle exige toutefois une prise de conscience accrue. Elle ne peut être dissociée d’une réflexion et d’une action portant sur la responsabilité (les personnes, la planète, la prospérité), la confiance et l’éthique – et, plus largement, la gouvernance. C’est sur le thème de la gouvernance que j'interviendrai, en m’appuyant sur les retours d’expérience recueillis sur le terrain, notamment auprès de plus de 150 C-Levels. Je délivrerai les principales conclusions de ces retours et de notre livre blanc sur la gouvernance de l'IA à 2030. Avec la contribution d’AdVaes et d’organisations travaillant sur cette question.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4592,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "10:35",
  "end": "11:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "The Digital Quiz, par La Fresque du Numérique",
  "speakers": [
   {
    "name": "Jérôme Vigogne",
    "title": "Consultant",
    "company": "ACT4CHANGE"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Workshop on AI environmental impact; outside the rubric.",
  "abstract": "Workshop lead in French, translated in-time in English. You use AI without knowing its environmental impacts? Discover impacts of AI through the digital life cycle. The Digital Quiz is a 20’ demo of our 3h workshop The Digital Collage. +180 000 people took part — why not you? — Vous déployez l’IA sans connaître ses impacts environnementaux ? Venez découvrir les impacts de l’IA au travers du cycle de vie du numérique. Le Quiz Numérique est une démonstration de 20’ de l’atelier de 3h La Fresque du Numérique. +180 000 personnes sensibilisées, pourquoi pas vous ?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4574,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "10:35",
  "end": "11:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Trusting the “Green” in Tech: How ISO 14064 Third-Party Verification Combats Greenwashing",
  "speakers": [
   {
    "name": "Melanie Blackmore",
    "title": "CEO and Founder",
    "company": "Carbonology"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Third-party verification of carbon claims; disclosure no facet reads.",
  "abstract": "Tech and AI companies are publishing carbon numbers, but a published number isn’t a credible one. The European Commission found 53% of green claims were vague, misleading, or unfounded. With ECGT applying since September 2026 and CSRD, UK SRS and SB 253 expecting assured data, self-reported emissions are a growing liability. This talk explains how independent verification under ISO 14064-3 works, common data gaps, and how to make emissions data audit-ready and reusable across disclosures. It also asks where and how AI can ease verification and where human independence must remain.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4829,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "From AI Models to AI Systems",
  "speakers": [
   {
    "name": "Gilles Closset",
    "title": "Global AI Ecosystem Leader",
    "company": "OVH"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; AI models to systems, no API surface stated.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4536,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Planned Obsolescence & Ecological Design",
  "speakers": [
   {
    "name": "Laetitia Vasseur",
    "title": "Déléguée Générale",
    "company": "Halte à l'Obsolescence Programmée"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Planned obsolescence in IT; outside the rubric.",
  "abstract": "Presentation of the NGO's missions and expertise. Key note on the importance of planned obsolescence on IT and its key role in planning for a more sustainable path.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4827,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "How to enforce structured-outputs for Vision Langage Models",
  "speakers": [
   {
    "name": "Sonia Tabti",
    "title": "AI and Computer Vision specialist",
    "company": "Sonia Tabti"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; enforcing structured outputs is schema-constrained responses.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4649,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Securing the Supply Chain : A Practical Guide",
  "speakers": [
   {
    "name": "Isabelle Mauny",
    "title": "Field CTO",
    "company": "Choreo Labs, formerly WSO2"
   }
  ],
  "facets": [
   "operational_transparency",
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "Software supply chain security practice.",
  "abstract": "Over the years, software components morphed from being developed in-house to being consumed from outside sources. While this greatly accelerates productivity, it also greatly expands the enterprise attack surface. In this session, we share practical approaches used by large enterprises we work with. We will include takeaways from their successes and failures, illustrated by real-life breaches that occurred via supply chain weaknesses, including NPM, MCPs, or Docker images.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4635,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Agentic AI Is Inevitable. How Do You Govern It Safely Without Losing Control of Costs?",
  "speakers": [
   {
    "name": "Anuj Kumar",
    "title": "Principal Product Manager",
    "company": "Axway"
   },
   {
    "name": "Jose Monteagudo",
    "title": "Principal Pre-Sales Architect",
    "company": "Axway"
   }
  ],
  "facets": [
   "agent_readiness",
   "access_clarity",
   "accountability"
  ],
  "agent_readiness_dimensions": [
   "agent_identity_declared",
   "rate_limit_signal"
  ],
  "why": "Who uses which AI and API, at what cost, under which rules.",
  "abstract": "Agentic AI is driving a surge in interactions across LLMs, APIs, and tools. The challenge is no longer simply accessing AI, but knowing who uses what, at what cost, and under which rules.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4557,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Streaming's Energy Use. Watts, Methods, and What the Industry Needs to Measure Next",
  "speakers": [
   {
    "name": "Benjamin Schwarz",
    "title": "President",
    "company": "Greening of Streaming"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Measured streaming energy use; outside the rubric.",
  "abstract": "Most streaming energy figures are modeled. Greening of Streaming spent 4 years measuring instead: real devices, codecs, and traffic, metered at the socket. Across a 128-fold bitrate range, a streaming box moved under half a watt while the TV drew ~50 W. We cover where the watts go, the methodology that makes numbers defensible, and open questions: is per-GB network energy a measure of efficiency or just pipe fill? Why do some devices draw significantly more with newer codecs? Where should the home boundary sit? Green IO attendees will see a new GoS dataset first.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4648,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "API Developer Experience as Strategic Lever",
  "speakers": [
   {
    "name": "Miriam Greis",
    "title": "Principal Integration Architect.",
    "company": "adorsys GmbH + Co. KG"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "API developer experience as a strategic lever.",
  "abstract": "Developer Experience (DX) has long become a critical success factor for API initiatives. Whether you are providing or consuming APIs: friction in development slows adoption, increases support effort, and in the worst case can cause integrations to fail. In this talk, we focus on what DX means, especially in the context of APIs and how DX for providers and DX for consumers are linked. Based on Team Topologies, we explore how a platform team can improve DX for both API providers and consumers. This includes topics such as consistent design, documentation, maintenance, API discovery, and the lifecycle of APIs. API Developer Experience is not a \"\"nice to have\"\", but a strategic lever for platform growth and the success of the ecosystem.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4656,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Your Agents Are Running the Organisation ,Who's Watching Them?",
  "speakers": [
   {
    "name": "Kelly Policarpo",
    "title": "Founder & CEO",
    "company": "Lumo AI Compliance & Advisory"
   }
  ],
  "facets": [
   "accountability",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "agent_identity_declared"
  ],
  "why": "Oversight of agents operating the organisation.",
  "abstract": "AI agents have stopped assisting and started operating. They run workflows, make decisions, and execute tasks that used to require a person. In many organisations, humans are quietly becoming the minority in functions they once owned outright. This talk examines what governance looks like once autonomous AI becomes the workforce, not just the tool. Drawing on European regulatory frameworks and experience shipping AI systems in HR Tech, Legal Tech, and Health Tech, I introduce a practical framework for what human presence an organisation still needs: who holds authority, who's accountable, where humans still participate, and what continuity requires when today's process owners may not be there tomorrow. Why it matters: Most agentic AI conversations focus on capability. Almost none focus on governance, on what breaks when agents act without anyone watching. Teams shipping agents into production are answering the \"\"who's in charge\"\" question today, often without realising it. This talk gives builders and product leaders a concrete lens to audit their own systems before a regulator, customer, or incident forces the question.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4344,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "How Back Market scaled from 0 to 320 engineers on Coding Agents",
  "speakers": [
   {
    "name": "Nicolas Martignole",
    "title": "Principal Engineer",
    "company": "Back Market"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Rolling out coding agents to engineers; no API surface.",
  "abstract": "In this talk, I present how I scaled from 0 to 320 users on Claude Code, and now Open Code. I'll share what worked for us, our successes... and our failures. We learned the hard way. I'll also share my vision and the impact on the daily work for an engineer, and for non-technical contributors at Back Market.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4566,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Sustainable, Sovereign AI: Do companies really give a sh*t ?",
  "speakers": [
   {
    "name": "James Martin",
    "title": "Founder",
    "company": "BetterTech"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Small sovereign models for sustainability; outside the rubric.",
  "abstract": "We know the theory. Smaller AI models are not only less impactful; you can also host them wherever you want, and keep them and their data for as long as you want. So they're great for sustainability and sovereignty; not to mention hundreds of times cheaper, or even free if run locally. But how many companies really care about making their AI usage more frugal and sovereign? More than you think! BetterTech's James Martin shares real-life insights gained since launching his \"Frugal AI\" training course, with GreenIT.fr, in September 2025.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4728,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Explaining GraphQL Federation to my non-technical wife",
  "speakers": [
   {
    "name": "David Stutt",
    "title": "Senior Founding Engineer",
    "company": "WunderGraph"
   }
  ],
  "facets": [
   "contract_quality"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; GraphQL federation explained.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4642,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "11:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "Offensive Security in the AI Era: New Attacks, New Reflexes",
  "speakers": [
   {
    "name": "Noureddine Taher",
    "title": "Test engineer",
    "company": "BNP Paribas"
   }
  ],
  "facets": [
   "operational_transparency",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI-enabled attacks and security posture.",
  "abstract": "Artificial intelligence is rapidly transforming the cybersecurity landscape for both attackers and defenders. This presentation explores new AI-enabled attack techniques, their impact on organizations, and the challenges they pose for security teams. Through concrete examples and lessons learned, we will also examine how to adapt offensive security approaches and strengthen organizations’ ability to detect and anticipate these emerging threats.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4830,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:20",
  "end": "12:15",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "The Product Design Assembly Line",
  "speakers": [
   {
    "name": "Ivan Dalmet",
    "title": "CTO",
    "company": "Peinture"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; product design process, no API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4541,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:45",
  "end": "12:10",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "La seconde vie des équipements IT : un levier environnemental et économique",
  "speakers": [
   {
    "name": "Helene Lieutenant",
    "title": "Head of End User Support",
    "company": "PartnerRe"
   },
   {
    "name": "Etienne Viemont",
    "title": "Fondateur et gérant",
    "company": "ARTO"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Second life of IT equipment; outside the rubric.",
  "abstract": "Comment transformer les parcs informatiques en fin de vie en ressources ? Évoquer la quantité de matériel dormant dans les entreprises, les possibilités de réemploi et les arbitrages entre réparation, reconditionnement et recyclage.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4573,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:45",
  "end": "12:10",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Conscious Use of LLMs: Efficiency and Sustainability in Practice",
  "speakers": [
   {
    "name": "Nicolas Fantoni",
    "title": "Platform Engineer",
    "company": "Imola Informatica S.P.A."
   },
   {
    "name": "Riccardo Soro",
    "title": "DevOps Consultant",
    "company": "Imola Informatica"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Efficient, sustainable LLM use; outside the rubric.",
  "abstract": "Generative artificial intelligence has enormous potential, but also a significant environmental cost. In this talk, we will see how to make the use of Large Language Models more efficient and sustainable through techniques such as quantization, pruning, and distillation. Through concrete use cases, we will explore the advantages of model optimization: we will analyze the differences in terms of size, speed, resource consumption, and response quality. We will also discuss models that are already “green by design” and open source tools for working with smaller but still effective models.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4593,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "11:45",
  "end": "12:10",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Adecco × Sopht : retour d’expérience sur une démarche Green IT qui tient la route",
  "speakers": [
   {
    "name": "Julien Bryard",
    "title": "Head of IT Infrastructure & Operations Europe",
    "company": "Adecco"
   },
   {
    "name": "Isabel Jaquenoud",
    "title": "Delivery & Customer Success Manager",
    "company": "Sopht"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Green IT case study; outside the rubric.",
  "abstract": "Comment faire durer le Green IT quand le green n’est plus la priorité ? Julien Bryard (Adecco) et Isabel Jaquenoud (Sopht) reviennent sur ce qu’Adecco a mesuré, les actions engagées et la valeur business obtenue. Ils terminent sur le prochain défi : l’IA, et pourquoi il ne faut pas se contenter des chiffres des big tech.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4695,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "From API Traffic to Agent Execution: Why AI Agents Need a Different Gateway",
  "speakers": [
   {
    "name": "Gbadebo Bello",
    "title": "Senior Developer Relations Advocate",
    "company": "Postman"
   }
  ],
  "facets": [
   "agent_readiness",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "rate_limit_signal",
   "idempotency",
   "agent_identity_declared"
  ],
  "why": "Agents retry and act autonomously, so gateways need different controls.",
  "abstract": "AI agents do not behave like traditional API clients. They choose tools dynamically, call models and services in changing sequences, carry context across steps, and can retry or act autonomously. That makes the familiar request-by-request controls of an API gateway necessary, but no longer sufficient. This talk explores the shift from managing API traffic to governing agent execution. We'll look at the identity, authorization, tool discovery, context, reliability, cost, and observability challenges that emerge when agents work across LLMs, MCP servers, APIs, and other agents. Then, through a live Banking investigation demo, we'll compare a direct agent integration with a gateway-mediated path to show what changes when control is applied across the whole execution flow, not just at a single endpoint.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4336,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Your Next API Consumer Is Not Human: Designing APIs for AI Agents",
  "speakers": [
   {
    "name": "Hussein Loubani",
    "title": "AI Researcher",
    "company": "CIAD"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "operation_distinctiveness",
   "error_semantics",
   "openapi_examples",
   "spec_presence"
  ],
  "why": "Designing APIs an agent picks, fills and acts on without a human.",
  "abstract": "Your next API consumer is not a developer reading your documentation but an AI agent that picks a tool, fills in the parameters on its own, and acts on whatever your API returns without a human in between. Most APIs were never designed for this kind of caller, and the cracks show fast. I learned this the hard way while building agentic AI systems and connecting my research workflows through the Model Context Protocol. Whenever the agent failed, the model was rarely the problem, because the real issue lived in the API: a vague tool description sent it to the wrong endpoint, a loose schema let it guess instead of failing cleanly, and an error written for humans gave it nothing to recover from. This talk shows what changes when the consumer is a machine, and why it matters now: agents are already calling your APIs whether you planned for them or not. You leave with a practical checklist for making yours agent-ready, drawn from real failures rather than vendor slides.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4856,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "From API Guidelines to Governance Skills",
  "speakers": [
   {
    "name": "Daniel Kocot",
    "title": "Principal API Consultant",
    "company": "adorsys"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "agent_skills"
  ],
  "why": "API guidelines turned into governance skills in the daily workflow.",
  "abstract": "API governance rarely fails because organizations lack guidelines. It fails because those guidelines are often too far away from the daily work of service teams. They exist in documents, review meetings, and internal portals, but they are not always present when API design decisions are actually made. This talk builds on spec driven API design and looks at how Skills can help organizations bring API governance closer to development. Instead of treating API guidelines as static documentation, teams can turn design principles, naming conventions, review criteria, security expectations, and domain standards into reusable Skills that support developers while APIs are being designed and evolved. The session shows how Skills can assist with OpenAPI reviews, resource naming, consistent error models, breaking change awareness, and better design feedback. The goal is not to replace architects, API platform teams, or governance boards. The goal is to make their knowledge available earlier, more consistently, and closer to the workflow where API quality is created.** The talk closes with a practical model for moving from static guidelines to living governance capabilities, and with a critical reminder: Skills only become governance when they have ownership, versioning, feedback loops, and accountability.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4730,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "JSON Schema Standardization and Versions",
  "speakers": [
   {
    "name": "Lisa Dusseault",
    "title": "CTO",
    "company": "Data Transfer Initiative"
   }
  ],
  "facets": [
   "contract_governance",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [],
  "why": "JSON Schema standardization at the IETF and versions.",
  "abstract": "One of the most important tools to AI and APIs is now the subject matter of an active standardization Working Group at the IETF. Yet there's no reason to cut off innovation and extensions. How is this going to work? Who can contribute? (hint: everyone). Get informed, maybe even get involved.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4505,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "Zero Trust and Agentic AI Security",
  "speakers": [
   {
    "name": "Rob Wilson",
    "title": "",
    "company": "Broadcom"
   }
  ],
  "facets": [
   "agent_readiness",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "delegated_identity",
   "agent_identity_declared",
   "auth_clarity"
  ],
  "why": "Zero Trust for non-human agent identities.",
  "abstract": "In 2026, Non-human identities outnumber human employees 144-to-1, rendering traditional Zero Trust ineffective for autonomous AI agents. This session covers the transition to Agentic Zero Trust.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4935,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "When AI Agents Meet Your API Gateway",
  "speakers": [
   {
    "name": "Iliass Ayaou",
    "title": "Chief Research Officer",
    "company": "GroupBees"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "mcp_tool_quality",
   "operation_distinctiveness"
  ],
  "why": "Gateway APIs exposed as MCP tools, tested with 88 tools.",
  "abstract": "Your API gateway can expose every API as an MCP tool. But can AI agents use them reliably? We put 88 tools to the test with open-weight models. Surprising failures, hard lessons, and 3 checks to run.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4339,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "What Did the Agent Know, and When? Auditable Memory and API Guardrails for Regulated AI",
  "speakers": [
   {
    "name": "Abhijit Dey",
    "title": "Sr.Director FS Advisory",
    "company": "PWC"
   }
  ],
  "facets": [
   "regulatory",
   "accountability",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "reversibility_documented"
  ],
  "why": "Auditable agent memory and API guardrails in regulated industries.",
  "abstract": "Rule-based systems were auditable by default. You could read the code and know exactly what happened. AI agents break that assumption — they reason, adapt, and carry context forward in ways nobody wrote down in advance. Which leaves regulated industries with an uncomfortable question: when an examiner asks why a customer was declined last March, can you actually answer? This talk argues that two architectural layers make agents governable, and neither of them is the model. APIs are the action guardrail — the deterministic boundary between what an agent decides and what actually happens to a customer's money. The agent is allowed to think. The APIs define what it's allowed to do. Memory is the evidence guardrail — and it's the one almost everyone skips. An agent's decision is a product of what it knew at that moment: retrieved context, prior interactions, the policy version in force. Memory that isn't versioned, scoped, and reconstructable is a compliance liability wearing a productivity costume. You'll leave with a concrete architecture pattern, a reconstruction test for your own systems, and the questions to ask before your first agent reaches production.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4581,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:10",
  "end": "12:35",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Frugal by Design: How Wiley measures and manages the impact of their AI products",
  "speakers": [
   {
    "name": "Boris Gamazaychikov",
    "title": "Co-Founder and CEO",
    "company": "Sustainable AI Group (SAIG)"
   },
   {
    "name": "Laura Gunn",
    "title": "Senior Director, Sustainability and ESG",
    "company": "Wiley"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Measuring AI product impact; outside the rubric.",
  "abstract": "How can AI products do more with less? Wiley combines deterministic code, small models and frontier LLMs to build frugal AI research tools. Using Sustainable AI Group’s CLEER methodology, Wiley estimated that ARI’s Landscape View used about 88% less generative-AI energy and carbon than an equivalent LLM-only approach. Learn how the design and measurement work, how results inform product decisions, and get a first look at an end-to-end literature review case study comparing the full technology stack with a frontier-LLM workflow.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4598,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "12:10",
  "end": "12:35",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "AI, ally or enemy of the planet?",
  "speakers": [
   {
    "name": "Rosendo Mañas Faura",
    "title": "Partner Consulting & Head of Partnerships",
    "company": "Resilio"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI energy, carbon and water transparency; outside the rubric.",
  "abstract": "Today, no one really knows the energy consumption, carbon emissions or water consumption of AI models. This lack of transparency throughout the ICT value chain poses a real problem to businesses. Resilio is leading a collaborative study, involving academics, private and public organizations, and institutions, with the aim of raising awareness of the environmental impacts of AI, quantifying them, and turning this environmental information into a management and governance tool.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4538,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:10",
  "end": "12:35",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "L'inspirothèque pour un numérique plus écologique",
  "speakers": [
   {
    "name": "Thomas Thibault",
    "title": "Designer - Researcher",
    "company": "Designers Éthiques - Limites Numériques"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Catalogue of ecological digital design; outside the rubric.",
  "abstract": "Quelles formes, fonctions et techniques d'un numérique s'inscrivant dans les limites planétaires ? L'équipe Limites Numériques a documenté et classé plusieurs centaines d'inspirations et leviers dans une \"inspirothèque\" en ligne. De la conception de services pour dispositifs obsolètes, à la création d'un numérique saisonnier, en passant par un design sensible pour représenter l'empreinte de nos appareils, nous explorerons les vastes possibilités créatives pour ouvrir d'autres voies d'un numérique alternatif.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4643,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "The Holiday Checkout Has No Human",
  "speakers": [
   {
    "name": "Tim Chang",
    "title": "VP of Application Security Products",
    "company": "Thales"
   }
  ],
  "facets": [
   "agent_readiness",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "agentic_commerce",
   "agent_identity_declared",
   "consent_identity"
  ],
  "why": "Telling trusted shopping agents from bots in API journeys.",
  "abstract": "Autonomous shopping agents will reshape holiday commerce and amplify fraud. This session reveals how to secure API journeys by separating trusted automation from bot abuse.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4828,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 1/2"
  ],
  "title": "From Motion Analysis to Agentic Video Understanding",
  "speakers": [
   {
    "name": "Yannis Tevissen",
    "title": "Head of Science",
    "company": "Moments Lab"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "History of video understanding; no API surface.",
  "abstract": "Video is the largest and least structured kind of data on the internet. This talk traces how machines learned to watch: from optical flow and early action recognition, through CNN-based video classification, to today's vision-language models that can describe long videos and answer questions about them. It then looks at what comes next, when video understanding stops being a single model call and becomes an agent. That agent chooses which frames to look at, combines vision, speech and text, and searches across whole video libraries to answer a question. Drawing on research at Moments Lab, the talk covers where current models still fall short (temporal awareness, long videos, compute cost, bias) and what it takes to build reliable agentic video systems in production.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4508,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Designing for developers who've changed how they build: lessons from Azure Developer CLI data",
  "speakers": [
   {
    "name": "Kristen Womack",
    "title": "Principal Product Manager",
    "company": "Microsoft"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "agent_skills"
  ],
  "why": "DX when coding agents mediate API use, from CLI data.",
  "abstract": "The way developers interact with APIs has changed in a way that's not going back. Not because developers are gone (they're not) but because their relationship to your API is increasingly mediated by a coding agent. And DX has to catch up to that reality. This talk maps two new developer personas that are already in your telemetry, whether you've named them or not: the developer building an agent that consumes APIs autonomously, and the developer building anything, assisted by coding agents making calls on their behalf. Drawing on growth data from the Azure Developer CLI and experience leading DX for some of the world's largest API platforms, this is a practitioner's view of what these two personas need today, and what breaks when we design for neither of them.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4944,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Composable Architecture for AI Systems: APIs, MCP, Skills, Sub-Agents, and Workflows",
  "speakers": [
   {
    "name": "Erik Wilde",
    "title": "Chief Evangelist",
    "company": "WunderGraph"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "agent_skills"
  ],
  "why": "Composing APIs, MCP, skills, sub-agents and workflows.",
  "abstract": "AI is a powerful tool, but it is not the only tool, and most systems worth building are hybrids of AI and non-AI parts. That line between them doesn't sit neatly between domains. It runs inside every domain, and it should move as you learn. This talk looks at the mechanisms for connecting across that line: APIs, MCP, agent skills, and sub-agents, as well as at workflows as a way to move the line itself: when an agent keeps rediscovering the same business process, that process belongs in deterministic execution, not in a prompt. You'll leave knowing which mechanism belongs where, and how to keep shifting work out of AI as your system settles.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4652,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "AI Adoption Success Secrets.",
  "speakers": [
   {
    "name": "Omid Eidivandi",
    "title": "Solution Architect",
    "company": "Groupe La Centrale"
   }
  ],
  "facets": [
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Shaping and securing enterprise AI adoption.",
  "abstract": "In an era where AI adoption in enterprises is accelerating at an unprecedented pace, companies must define a clear and sustainable path to success. This talk will explore how to shape and secure that adoption journey, highlighting the most critical factors and key decisions organizations need to consider along the way.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4831,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "Retaining governance of your projects by deploying Open-Weight Models",
  "speakers": [
   {
    "name": "Matthieu Boussard",
    "title": "Head of R&D",
    "company": "Craft AI"
   }
  ],
  "facets": [
   "open_source",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; self-hosting open-weight models is a sovereignty choice, not a published API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4504,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Beyond the Demo: Running AI Agents in Production to Modernize Enterprise APIs",
  "speakers": [
   {
    "name": "Vedran Cindrić",
    "title": "Founder & CEO",
    "company": "Treblle"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Agents refactoring a legacy API landscape in production; the contracts are what they rewrite.",
  "abstract": "This talk is built entirely on first-hand experience deploying AI Agents inside a large enterprise to refactor and modernize a legacy API landscape. Not in a sandbox, not in a lab, but in production, with real constraints, real data, and real consequences. I'll walk through what it actually takes: working within AWS Bedrock with sovereign data isolation requirements, spinning up sandboxed containers on AWS ECS in under 20 milliseconds, designing and optimizing agent Skills for code-level API refactoring, and integrating MCP servers for GitHub and Treblle to give agents the context they need to act intelligently. You'll leave with a clear picture of the architecture decisions that worked, the assumptions that didn't hold, and the patterns that made agentic modernization viable at enterprise scale.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4365,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Why is an AI gateway no longer optional",
  "speakers": [
   {
    "name": "Saad Amal",
    "title": "Software Engineer",
    "company": "ADEO"
   }
  ],
  "facets": [
   "access_clarity",
   "operational_transparency",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "rate_limit_signal",
   "auth_clarity"
  ],
  "why": "A stolen key and a runaway bill: quotas, key protection and cost controls at the gateway.",
  "abstract": "While scrolling on Reddit, I found this interesting (yet frightening) post: “$82,000 in 48H from a stolen Gemini API key. My monthly usage is $180. Facing bankruptcy.” This story is a powerful reminder of why an AI gateway is a critical component in ADEO's AI governance. In this talk, we will dive into the 'why' and 'how' of building an AI Gateway. We will share the inside story of how ADEO tackled this challenge to create a secure, multi-provider entry point for our developers, while lifting the curtain on the remaining challenges we still have to overcome.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4731,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "12:20",
  "end": "12:45",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Events as Contracts, Events as Data: Our AsyncAPI Journey",
  "speakers": [
   {
    "name": "Ignacio Castillejos",
    "title": "Staff Engineer",
    "company": "The New York Times"
   }
  ],
  "facets": [
   "contract_quality",
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "AsyncAPI as the contract between every event producer and consumer through a broker migration.",
  "abstract": "Migrating an event-driven architecture isn't just a change of broker. It means renegotiating contracts between every producer and consumer. In this talk, we share our journey from a legacy EDA built on GCP Pub/Sub and Avro to one built on AWS EventBridge, where every event follows the CloudEvents standard and the entire event API is defined by an AsyncAPI spec. We'll also show how those same events now power our data products. We'll cover: Why we changed. Routing messages on our legacy stack was hard, the setup carried too much infrastructure to maintain, and Avro depended on a schema registry we didn't control. Contract-first events with AsyncAPI. We govern our event schemas contract-first, the same way we govern our REST APIs with OpenAPI, with the spec as the source of truth. We'll show how that replaced the external schema registry and what the workflow looks like day to day. From events to data products. Because every event has a well-defined contract, the same events that drive our services land in our data lakehouse, where they're transformed into data marts and continuously monitored for quality. Attendees will leave with practical lessons on using AsyncAPI to make events a reliable foundation for both services and data products.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4958,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "13:30",
  "end": "14:00",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Partner's Village Booth #20"
  ],
  "title": "Offical Launch of FOSTBeat",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; a product launch with no stated API content.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4594,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "The journey behind the world's first sustainability certification for cloud resources: From concept to application",
  "speakers": [
   {
    "name": "Barton Finn",
    "title": "Project Manager - TCO Certified Cloud",
    "company": "TCO Development"
   },
   {
    "name": "Simon Shillaker",
    "title": "Environmental impact monitoring, Exoscale",
    "company": "Exoscale"
   }
  ],
  "facets": [
   "outside",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "A sustainability certification for cloud resources; a compliance badge the rubric does not read.",
  "abstract": "What does it take to turn sustainability criteria into real-world action? A year ago at Green IO Paris, we introduced the work behind TCO Certified Cloud, the world's first sustainability certification for cloud resources. With the criteria published in 2026, the focus has shifted from development to implementation. Together with Exoscale, TCO Development will share the journey from developing the criteria to implementing the certification, drawing on experiences from getting cloud resources certified. We'll explore challenges, lessons learned and opportunities for local cloud providers.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4655,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "AI Agents Evaluation with synthetic data",
  "speakers": [
   {
    "name": "Louis Jourdain",
    "title": "AI Engineer",
    "company": "ChapsVision"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "error_semantics"
  ],
  "why": "Evaluating multi-step agent reliability with synthetic data; agent-side testing, not the API surface.",
  "abstract": "As enterprises transition from reactive chatbots to proactive ReAct-based agents, a critical bottleneck has emerged: the Reliability Gap. Current benchmarks show that while top-tier models excel at single-turn reasoning, their success rate plummets by up to 75% in long-horizon, multi-step workflows. We can no longer rely on static test suites or manual \"\"vibe checks\"\" to evaluate agents that operate in dynamic, ever-changing environments.This talk introduces a novel framework for the automated generation of synthetic, state-aware evaluation datasets. Moving beyond simple Q&A, our system utilizes a multi-agent architecture to synthesize complex dual-control scenarios, inspired by the rigor of tau bench ( https://arxiv.org/abs/2506.07982 ) where both the user and the agent possess tools to modify a shared world state. To deploy agents in mission-critical sectors like Telecom, Finance, or Healthcare, we need a \"\"Generic Evaluation Engine\"\" that can: -Map custom tool definitions into verifiable state-transition trees. -Generate adversarial \"\"User Personas\"\" that stress-test policy adherence and edge-case recovery. -Implement metrics to measure consistency over stochastic drift.We will demonstrate how this \"\"Mirror World\"\" approach allows teams to identify cascading failures before they hit production, turning the \"\"messy reality\"\" of agent evaluation into a structured, engineering discipline.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4366,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "Securing and Industrializing AI Adoption at Scale",
  "speakers": [
   {
    "name": "Nejmeddine Ben Ouarred",
    "title": "CIO",
    "company": "GroupBees"
   },
   {
    "name": "Youssef Guenoun",
    "title": "Co-Founder and CTO",
    "company": "GroupBees"
   }
  ],
  "facets": [
   "agent_readiness",
   "operational_transparency",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity",
   "rate_limit_signal"
  ],
  "why": "A control layer between LLMs and the information system: access, security and quotas.",
  "abstract": "The adoption of AI and agents promises an efficiency revolution, but it raises a major challenge: how do we transition from an isolated prototype to a secure, industrial-grade platform? Without a control layer, opening LLMs to the information system exposes the enterprise to security risks, cost overruns, and non-compliance. In this talk, GroupBees shares its vision and expertise on the architecture of agentic platforms and the emergence of AI Gateways as a core pillar of AI-driven architectures. Agenda: Genesis of AI Gateways: From API Management to the AI Gateway Governance & Security: The AI Gateway as a \"firewall\" (granular authentication, content filtering, prompt injection prevention) Operational Control: Quotas, rate limiting, and traceability to manage costs and automated decisions Lessons Learned & Vision: Key success factors observed with our clients for deploying AI with total security Target Audience: CTOs, Architects, Product Owners, and Technical Leaders looking to transform the excitement around AI into a stable, governed production platform.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4324,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Beyond REST: Evolving Arazzo for Real-World Agentic Workflows",
  "speakers": [
   {
    "name": "Naresh Jain",
    "title": "Founder & CEO",
    "company": "Specmatic"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "event_surface_described"
  ],
  "why": "Arazzo workflows spanning OpenAPI, AsyncAPI, gRPC and GraphQL for agents.",
  "abstract": "A real-world business workflow rarely stays within the boundaries of a single API style. An order might begin with an HTTP API described using OpenAPI, trigger an event described using AsyncAPI, invoke a Protobuf RPC service, query data through GraphQL, and eventually receive an asynchronous notification. Agentic applications make this even more important: an agent needs to understand not just individual API operations, but how capabilities across heterogeneous systems can be composed into a meaningful workflow. Arazzo was created to describe sequences of API calls and the dependencies between them. But what happens when the next step isn’t another OpenAPI operation? In this talk, we’ll look at how Arazzo is evolving to support heterogeneous API interaction models. Drawing on the work to introduce AsyncAPI, Protocol Buffer RPC and GraphQL support into the Arazzo Specification, we’ll explore some of the surprisingly difficult design questions that emerge when REST-style request/response assumptions disappear. How do you represent send and receive operations in an event-driven system? What does a workflow step mean when an RPC is server-streaming or bidirectional? How should GraphQL queries, mutations and subscriptions participate in a workflow? And how can we add these capabilities without turning Arazzo into a collection of protocol-specific special cases? We’ll use these questions to explore a broader idea: Arazzo can become a protocol-independent description of how API capabilities are composed to accomplish an outcome. That has implications beyond API documentation and testing. For AI agents, an Arazzo workflow can provide something increasingly valuable: an explicit, machine-readable description of which capabilities to invoke, in what sequence, how data flows between them, and what successful execution looks like. Rather than asking agents to infer workflows from a collection of disconnected API descriptions, we can start making the workflow itself part of the contract.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4507,
  "day": "2026-12-02",
  "type": "Panel discussion",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Boomi Panel: State of the API Economy 2026",
  "speakers": [
   {
    "name": "Markus Müller",
    "title": "Global CTO - API Management",
    "company": "Boomi"
   },
   {
    "name": "Mark Boyd",
    "title": "Director",
    "company": "Platformable"
   }
  ],
  "facets": [
   "access_clarity",
   "discoverability"
  ],
  "agent_readiness_dimensions": [],
  "why": "State of the API economy report: a market-level read.",
  "abstract": "Join FOST Analyst Mark Boyd and Markus Müller, Global Field CTO at Boomi, as they discuss the results of the API Economy in the Age of AI: State of the Market Report (2026 Edition). They will cover key trends, forward-looking predictions, and practical insights on how API practitioners can prepare to build their customer base and adapt their infrastructure for 2027.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4361,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Why You (too) Need an (API) Style Guide",
  "speakers": [
   {
    "name": "Robert Lehmann",
    "title": "Staff Engineer (SRE)",
    "company": "Google Switzerland"
   }
  ],
  "facets": [
   "contract_governance",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "operation_distinctiveness"
  ],
  "why": "API style guides for naming and review consistency.",
  "abstract": "You’ve probably been in this situation: You thought long and hard about this new API of yours, but now it comes to naming — shock, horror, what will you do? Or, worse, you’re a reviewer and you’re not sure if the proposed API quite aligns with the status quo. That’s where API Style Guides come in and can eliminate (some of the) bikeshedding. In this talk, we will explore: What is an API Style Guide, and why could you possibly need one? What are the characteristics of a good style guide, and what constitutes a bad one? Some exemplary style guides for your to be inspired by (e.g. AIP, AEP) Where else can style guides help (it’s 2026, do I need to spell it out: AI) Robert will draw on his decade+ at Google where he was one of 3% of the company allowed to perform API reviews. His team maintained multiple style guides in order to scale reviews.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4547,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Vers un usage raisonnable et conscient de l'IA : l'empreinte IA de Back Market à 360°",
  "speakers": [
   {
    "name": "Aurélien Gervasi",
    "title": "Head of Engineering, AI",
    "company": "Back Market"
   },
   {
    "name": "Florian Pothin",
    "title": "AI-Carbon Data Scientist",
    "company": "Toovalu"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI carbon footprint at Back Market; sustainability disclosure the rubric does not measure.",
  "abstract": "Back Market a pour mission de rendre l'économie de la tech circulaire. Pour rester compétitifs face à l'industrie du neuf, nous utilisons l'IA, mais pas sans en estimer l'impact. Avec Toovalu, spécialiste de l'empreinte carbone de l'IA, nous avons établi un inventaire carbone et eau inspiré de l'analyse du cycle de vie. Il couvre tous nos outils d'IA internes, des agents de programmation aux bots de support. Découvrez comment cette démarche nous a permis de mieux comprendre nos usages et d'identifier les leviers pour réduire l'impact environnemental de l'IA.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4832,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Boris Gamazaychikov",
    "title": "Co-Founder and CEO",
    "company": "Sustainable AI Group (SAIG)"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4415,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "From Vendor Dependency to Open Source: A Journey in Trade-offs",
  "speakers": [
   {
    "name": "Sadique Akhtar",
    "title": "Software Engineer",
    "company": "Maersk GSC"
   },
   {
    "name": "Nitya Tripathi",
    "title": "Software Engineer",
    "company": "Maersk GSC"
   }
  ],
  "facets": [
   "open_source",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "SaaS versus open-source trade-offs for internal tooling; weak public API signal.",
  "abstract": "As software organizations face increasing pressure to optimize costs without compromising reliability, scalability, or developer experience, the choice between SaaS platforms and open-source alternatives is becoming more strategic than ever. In this talk, we share our journey of re-evaluating API gateway technology in response to changing cost considerations and the evolving open-source ecosystem. Starting with a SaaS-based gateway, moving to Kong OSS, and exploring Apache APISIX through a proof of concept, we will discuss the factors that shaped our decisions, the trade-offs we encountered, and the lessons learned along the way. The discussion goes beyond comparing gateway products. We will explore how to balance licensing and operational costs, engineering effort, platform capabilities, community contributions, long-term maintainability, and the risks of vendor dependency. We will also examine why adopting open source is not simply about reducing licensing costs, but about understanding the total cost of ownership and building the expertise needed to operate and evolve a platform sustainably. Our goal is to share a practical, experience-driven perspective on making technology choices in a cost-conscious environment—helping teams assess when open source makes sense, what trade-offs to expect, and how to make informed decisions without treating any single technology as a one-size-fits-all solution.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4880,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Your Portal Has Two Kinds of Visitors Now: Arazzo Workflows for Humans and Agents",
  "speakers": [
   {
    "name": "Pubudu Gunatilaka",
    "title": "Director, Head of Architecture - API Platform BU",
    "company": "Choreo Labs, formerly WSO2"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics",
   "agent_readiness",
   "discoverability"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "well_known_catalog",
   "agent_skills"
  ],
  "why": "Arazzo workflows published in the portal so humans and agents follow multi-call sequences.",
  "abstract": "OpenAPI tells an agent which operations exist. It does not tell it that placing an order takes four calls across three APIs, in a specific order, passing an identifier from step two into step four. Agents rediscover that sequence by trial and error, and they get it wrong in ways that are expensive to debug and risky to run in production. We built API Workflows into our developer portal on top of Arazzo. Each workflow pairs an Arazzo definition, which tells an agent exactly how the calls chain together, with a natural language agent prompt that tells it when reaching for this workflow is the right move. Agents turn out to need both: the mechanics and the intent. The more interesting half is publication. One workflow is rendered in the portal for human developers and emitted into llms.txt and api-workflows.md for agents, with visibility controlled independently per audience and a draft-and-publish lifecycle so nothing reaches either surface early. This talk covers what Arazzo gave us for free, what we had to build around it, and why publishing outcome-level workflows beats handing agents a catalog of endpoints. Key Takeaways: Why OpenAPI alone leaves agents guessing at multi-step business processes, and precisely what Arazzo adds Pairing a machine-readable Arazzo definition with natural-language invocation guidance, and why agents need both Publishing the same workflow to human and agent surfaces, including llms.txt and markdown, from one definition Governing agent-facing capability with independent per-audience visibility and a publish lifecycle Where Arazzo fit cleanly, where we had to extend around it, and what that suggests for the spec's evolution",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4578,
  "day": "2026-12-02",
  "type": "Panel discussion",
  "start": "14:00",
  "end": "14:40",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "How to Unlock the Double-Wins of AI and Cloud Sustainability",
  "speakers": [
   {
    "name": "William Tinney",
    "title": "CTO & Founder",
    "company": "Greenpixie"
   },
   {
    "name": "Jasper Behnken",
    "title": "Product Owner - Sustainable IT",
    "company": "ABN AMRO Bank"
   },
   {
    "name": "Mickaël Guevorkian Babaev",
    "title": "AI & Cloud Cost Control Executive",
    "company": "CloudHealth"
   }
  ],
  "facets": [
   "outside",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI and cloud sustainability with FinOps; only the finops piece touches Access Clarity.",
  "abstract": "Using real-life case studies, this session explores how enterprise customers can deliver on the promise of business value and address IT Sustainability goals. How to support the demands of Responsible AI, FinOps and Tokenomics programs. An opportunity to learn directly from practitioners.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4539,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:25",
  "end": "14:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Measuring AI's Footprint Together: The BoAmps Open Dataset",
  "speakers": [
   {
    "name": "Sarah Oury",
    "title": "Frugal AI expert",
    "company": "Sopra Steria"
   }
  ],
  "facets": [
   "outside",
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "An open dataset of field-measured AI energy use; providers rarely disclose it.",
  "abstract": "AI providers rarely disclose the energy cost of their models. Estimation tools exist, but they rely on lab conditions that don't reflect real-world usage. BoAmps, by the Boavizta organization, addresses this by collecting standardized, field-measured energy consumption data from anyone running AI workloads. The result is an open dataset that, for the first time, enables genuine comparison of models, architectures, and infrastructure choices based on empirical evidence. Ultimately, the goal is to turn that data into actionable insights for building frugal AI.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4565,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "14:25",
  "end": "14:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Why smart people disagree on sustainability?",
  "speakers": [
   {
    "name": "Romain Jacob",
    "title": "Senior Researcher",
    "company": "ETH Zurich"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Methodology disagreements in digital impact estimates.",
  "abstract": "Why do digital impact estimates vary by 10x, 100x, sometimes 1000x? How come respected and responsible scientists disagree on what to do with those estimates? How to interpret them? Answer: It boils down to the subtle methodological nuances and choices underpinning sustainable impact assessment. In this talk, I summarize those choices, illustrate how they matter in practice, and show how being unaware of them can easily lead you misunderstand what the numbers mean. Also, I explain why I used to be obsessed with consequential LCA, and why I was wrong.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4833,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "The AI Control Layer Europe is Missing",
  "speakers": [
   {
    "name": "Samy Melaine",
    "title": "CTO & Founder",
    "company": "Eden AI"
   }
  ],
  "facets": [
   "accountability",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; a European AI control layer, read from the title only.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4509,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Sell the Outcome, Not the Endpoint: API-First Storytelling",
  "speakers": [
   {
    "name": "Sandeep Bhaskar",
    "title": "Solutions Director",
    "company": "Wise"
   }
  ],
  "facets": [
   "access_clarity",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "Selling outcomes over endpoints to banks; how the API is presented to buyers.",
  "abstract": "Banks and financial institutions don't buy APIs — they buy outcomes: faster settlement, safer compliance, better customer experience. Yet most infrastructure vendors still pitch specs, uptime SLAs, and endpoint documentation, leaving the business case buried under technical detail. This talk makes the case for a different approach: selling API infrastructure through narrative storytelling, where every technical capability is reframed as a value story a CFO, Head of Payments, or Chief Risk Officer can act on. Drawing on real presales experience helping large banks and FIs adopt API-first payment infrastructure, this session shows how to build a compelling storyline that connects API-first design principles — modularity, composability, developer experience — to concrete business value: revenue growth, risk reduction, and speed to market. We'll cover how to structure a pitch around a customer's own pain points rather than a feature list, how to use storytelling frameworks to make complex integration decisions feel low-risk, and why API-first architecture itself is a value proposition, not just an engineering choice. Attendees will leave with a practical framework for translating technical capability into narrative, and for pitching infrastructure the way banks actually make buying decisions — around trust, outcomes, and long-term partnership, not just technical compliance. This matters because the next wave of banking innovation depends on infrastructure teams and vendors who can speak both languages fluently.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4732,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "From Validation to Semantic Guardrails: JSON Schema as the Contract  Layer for Agentic Application Development",
  "speakers": [
   {
    "name": "Jean-Pierre DIjcks",
    "title": "Product Manager",
    "company": "Oracle"
   },
   {
    "name": "Beda Hammerschmidt",
    "title": "VP JSON Development",
    "company": "Oracle"
   }
  ],
  "facets": [
   "contract_quality",
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "error_semantics",
   "spec_presence"
  ],
  "why": "JSON Schema as the semantic guardrail contract for agentic apps.",
  "abstract": "Agentic coding systems do more than generate code snippets. They assemble applications, call tools, wire APIs, query databases, transform data, and produce user-facing experiences. As these systems move from assistance to action, they need more than prompt-level guardrails. They need data-connected guardrails: machine-readable contracts that describe what data means, how it may be used, where it may flow, and which policies apply. This talk presents Project Delphi, an authoritative JSON Schema repository, and proposes a new vocabulary for extending JSON Schema with semantic and governance annotations. Project Delphi is a concrete implementation, but the larger goal is broader: to explore how the JSON Schema community defines a shared, interoperable vocabulary for data semantics, policy, and AI-era application contracts. The core idea is simple: JSON Schema should remain the structural contract for data, while the vocabulary adds the semantic layer needed by agents, tools, and generated applications. This can describe document types, field roles, sensitivity classifications, PII/internal-only markers, lineage, retention, masking requirements, and AI usage permissions, as well as non-functional requirements. We will explore how this changes the application development model. Instead of an agent merely generating code against loosely described APIs, it can consult authoritative schemas and use semantic annotations to make policy-aware decisions: which fields may be shown, logged, embedded, summarized, joined, exported, or passed to a tool. Applications generated or modified by agents can then project data safely and consistently according to schema-level policy, rather than relying on scattered hand-coded checks. The session will cover a concrete technical model for schema-backed guardrails: registry-managed schemas, semantic annotations, field-level policy enforcement, safe data projection, tool-call contracts, compatibility checks, and runtime validation. We will close with open design questions and an invitation for the JSON Schema community to collaborate on DSV as a standards-oriented vocabulary rather than a vendor-specific extension.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4644,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "Thales security white paper launch",
  "speakers": [],
  "facets": [
   "operational_transparency",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; a security white paper launch, read from the title only.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4503,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Where Kafka meets APIs: how VINCI Construction exposes and governs streaming data",
  "speakers": [
   {
    "name": "Yara Ayoub",
    "title": "Data Integration Analyst",
    "company": "VINCI Construction"
   },
   {
    "name": "Lucas Mundinger",
    "title": "Data Integration Project Manager",
    "company": "VINCI Construction"
   }
  ],
  "facets": [
   "contract_governance",
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "Exposing Kafka streams as governed HTTP APIs.",
  "abstract": "Discover how VINCI Construction bridges Kafka and HTTP, exposing event streams as governed APIs. Yara Ayoub and Lucas Mundinger share their architecture, lessons learned and future plans.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4694,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Beyond Faster Coding: Building the AI Software Factory",
  "speakers": [
   {
    "name": "Fred Plais",
    "title": "CEO",
    "company": "Upsun"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "AI software factory for delivery speed; no API surface in the abstract.",
  "abstract": "AI makes coding faster, but why aren’t roadmaps moving faster? Explore how AI software factories connect discovery, development, and operations to turn faster coding into faster delivery.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4629,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Scaling AI-Driven engineering without scaling risk",
  "speakers": [
   {
    "name": "Jean Burellier",
    "title": "Software Engineer",
    "company": "Mistral"
   }
  ],
  "facets": [
   "contract_governance",
   "operational_transparency",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "Security, compliance and governance keeping pace with AI-driven API output.",
  "abstract": "AI is changing how we build software. Development cycles are getting shorter, and the volume of code, services, and APIs reaching production is increasing. But as engineering teams move faster, traditional approaches to security, compliance, and governance struggle to keep up. How do we take advantage of AI-driven productivity without compromising reliability, security, or compliance? Drawing on experience building engineering platforms in highly regulated environments and working on large-scale AI infrastructure, this talk explores how we can move away from manual approvals and release gates toward automated guardrails built directly into our development workflows. We'll look at practical approaches to Compliance as Code, AI-assisted development, scalable CI/CD, self-service platforms, and continuous verification. We'll also discuss what happens when the bottleneck shifts from writing code to reviewing, validating, and deploying it. The goal isn't just to ship faster. It's to make sure our engineering platforms, security practices, and compliance processes can keep up as AI changes the way we build software.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4350,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "AI Agents FinOps: How to Master Your AI Budget",
  "speakers": [
   {
    "name": "Mariem Jabloun",
    "title": "AI Architect",
    "company": "YOWE"
   }
  ],
  "facets": [
   "access_clarity",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "rate_limit_signal"
  ],
  "why": "Controlling unpredictable token and tool-call costs of agents.",
  "abstract": "AI agents are moving from experimentation to production, but their costs are becoming increasingly unpredictable. Token consumption, model reasoning, agent loops, tool calls and autonomous workflows can make traditional cloud-cost controls ineffective. This talk presents a practical FinOps framework for AI agents: how to measure cost per agent/task, set budgets and guardrails, choose models intelligently, detect runaway workloads, and connect AI spend to business value. Attendees will leave with concrete techniques for controlling AI costs that saves them for bad surpirses.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4591,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:40",
  "end": "15:05",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "LeafCloud case: When does data-center heat reuse provide a net climate benefit?",
  "speakers": [
   {
    "name": "Hans Stokking",
    "title": "Sustainable ICT Lead",
    "company": "TNO"
   },
   {
    "name": "Louise Aubet",
    "title": "Head of R&D",
    "company": "Resilio"
   },
   {
    "name": "Sacha van Geffen",
    "title": "CTO",
    "company": "LeafCloud"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Data-center heat reuse climate benefit.",
  "abstract": "LeafCloud is a Dutch cloud provider operating small-scale datacenters, called LeafSites, in existing buildings such as apartment buildings. Heat from the IT equipment is used for domestic hot water, reducing gas use and CO₂ emissions. We present an ISO-compliant LCA (14040/14044 (LCA) & 14067 (PCF)) showing how CPU and GPU utilisation, electricity generation mix, displaced heat source, and daily and seasonal patterns affect annual emissions. We answer the question: is a LeafSite \"carbon-negative\", and under which scenarios?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4561,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "14:50",
  "end": "15:15",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Model your digital services with e-footprint",
  "speakers": [
   {
    "name": "Vincent Villet",
    "title": "Digital Services Modeling Expert",
    "company": "Boavizta"
   },
   {
    "name": "Michalina (Chwastyk) Cetin",
    "title": "Sustainable IT and Accessibility Consultant",
    "company": "Boavizta"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Modelling a digital service's environmental footprint.",
  "abstract": "Ecodesigning a service is less about \"what could I do?\" than \"what should I do first, and what impact would that have?\". Best practices tell you what's possible and measurement tells you what's happening today; e-footprint adds the missing leg: a model of your service, in real volumes and geographies, that lets you simulate a change before you make it and see where effort actually pays off. Come and learn to use this Boavizta open source tool through its web interface, and answer questions such as \"how would the footprint of my service evolve if I added generative AI in my user journeys?\".",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4543,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "14:50",
  "end": "15:15",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Empreinte environnementale de l'IA Gen en open source : chercheurs et entreprises ensemble pour le bien commun",
  "speakers": [
   {
    "name": "Caroline Darmon",
    "title": "CSR Director",
    "company": "Publicis"
   }
  ],
  "facets": [
   "outside",
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "An open alliance for GenAI footprint data.",
  "abstract": "Face aux manques de données disponibles sur cet impact, Publicis a créé avec les membres fondateurs AXA, Engie et le Groupe La Poste/ La Banque Postale, l'alliance \"GenAI footprint\" en association avec des chercheurs de Sustainable AI Group dont Sasha Luccioni et EcoLogits. L’objectif : quantifier et partager des données fiables sur l'empreinte CO2 des modèles GenAI pour la production de contenus. Ces données alimentent un outil open source conçu par Publicis Sapient qui inclut déjà les données pour estimer texte et vidéo permettant de partager des premiers résultats. FDJ United, Accor, L’Oréal, Orange et Renault Group ​sont également membres partenaires de cette alliance, qui reste ouverte à d’autres, pour continuer de financer les travaux en cours et à venir sur l’image, le son (…) et l’impact au delà du CO2.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4359,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Sovereign DevOps-Driven API Governance in Practice: Open Source and Agent-Ready in Hours",
  "speakers": [
   {
    "name": "Andrzej Jarzyna",
    "title": "API consultant",
    "company": "PZU"
   }
  ],
  "facets": [
   "contract_governance",
   "open_source",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Governance as linting in the DevOps pipeline with open-source tooling, agent-ready.",
  "abstract": "API governance is often discussed in terms of guidelines, standards, and central API catalogues. In many organisations, though, it is still too manual, too slow, or simply arbitrary. In this talk, we will show how to make API governance part of the DevOps process. Using open source and independent tools and CI/CD pipelines, we can automatically check whether APIs follow company standards, catch breaking changes early, keep APIs catalogued, and reduce drift between guidelines and what teams actually deliver. We will also show how to put in place a bare-bones, no-fuss governance setup in just a few hours, one that brings value quickly without becoming a big governance initiative. Finally, we will show how to treat AI agents as first-class citizens, so APIs are designed to be understandable and safe to use not only for developers, but also for agent-based consumers. We will present all of this in practice in a simulated environment.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4334,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "No Score, No MCP: Making APIs Earn Their Way Into the Agent Economy",
  "speakers": [
   {
    "name": "André Cedik",
    "title": "IT Architect",
    "company": "Kuehne+Nagel"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "spec_presence",
   "mcp_tool_quality"
  ],
  "why": "APIs must pass a score before getting an MCP server; close to the Agent Readiness rubric.",
  "abstract": "Every team wants an MCP server. Almost none of their APIs are ready for one. We're a platform team responsible for 200+ APIs we don't own and built by teams that don't report to us. For years we let them deploy and manage their APIs using our platforms' UI. This led to a complicated governance mess, and we had no lever to always force good API design onto them. In 2025 we've introduced a GitOps based deployment flow and things got interesting. And in 2026 agents arrived as the new consumers, and all of a sudden good API design was a necessity to make the agents understand and use the APIs. So we added an AI-readiness dimension to our API quality score — deterministic, static-analysis only, every violation traceable to a line in the spec and a named rule — and we're making that score a precondition for getting an MCP server. Not a block on what you have; a gate on what you want. That changes how teams respond entirely. This talk covers the mechanism and the mess: what makes an API genuinely readable by an agent rather than merely wrapped in one, how we scored that without starting an argument we'd spend a year losing, what had to be true about our deployment pipeline first, how teams actually reacted, and what we got wrong along the way.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4634,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Build your integrations for the business you're becoming - The integration platform you grow on",
  "speakers": [
   {
    "name": "John Bilay",
    "title": "Head of Technology",
    "company": "Alldotz"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Sponsor pitch for an integration platform.",
  "abstract": "In this session, we will show you one integration platform for people and AI agents that takes you from your first flows to the one your business can’t run without. Easy to start, sturdy from day one, so growing never means rebuilding. Alldotz is for teams whose integration needs grow faster than they can build. Start from a catalogue of 200+ connectors instead of a blank page, describe what you need and watch the flow appear on the canvas, so you build in a fraction of the time. Alldotz knows the systems around you: it holds their metadata and data models, so AI can generate the field mapping for you and all you do is review it. Then watch it work: simulate the flow with mock data, or dry-run it read-only against your real systems, and see every step light up as it runs. Governance keeps a to-do list of what needs your attention, from failing runs and policy breaches to deprecated functions, forgotten drafts, potential performance problems and suggested improvements to your flows and your integration landscape, so you hear about problems from Alldotz, not from your users. When a run fails, AI explains why in plain language, and self-healing proposes the fix for you to approve. And what your team builds is not just for people: expose any flow as a tool for your AI agents, so agents and humans use the same dependable integrations on one platform, with agent access protected and under your control. A landscape view shows how all your systems and integrations connect, and documentation is written for you by AI, ready for your team to refine. Less time spent building, mapping and troubleshooting means faster delivery and a quicker return on your integration investment, and growing never means rebuilding.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4638,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Govern APIs to Accelerate AI ROI",
  "speakers": [
   {
    "name": "Markus Müller",
    "title": "Global CTO - API Management",
    "company": "Boomi"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "mcp_server"
  ],
  "why": "Governing and exposing APIs as MCP tools; vendor framed.",
  "abstract": "Join Markus Müller to learn how an enterprise-grade platform enables teams to govern, monitor, productize, and safely expose their APIs as MCP-enabled, agentic tools.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4834,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "Building an AI Model Router: Practical Lessons and Designs",
  "speakers": [
   {
    "name": "Jason Davenport",
    "title": "Area Technical Lead",
    "company": "Google Cloud"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; model routing is LLM infrastructure, not a published API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4733,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "From Schema Registry to Agent Registry: Governing AI Discovery with the A2A Protocol",
  "speakers": [
   {
    "name": "Carles Arnal",
    "title": "Principal Software Engineer",
    "company": "IBM"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "agent_card",
   "well_known_catalog"
  ],
  "why": "An A2A agent registry carrying discovery and versioning governance to agents.",
  "abstract": "API governance solved discovery, versioning, and compatibility for REST services. Now AI agents face the same challenges. One team updates a prompt template and downstream agents silently break. Welcome to the API governance problem all over again but for agents. This talk shows how Apicurio Registry (CNCF sandbox) evolved to manage AI-native artifacts using the same governance patterns that worked for APIs: A2A Agent Cards for discovery, LLM Model Schemas for metadata, and prompt templates with version control. Live demo: agents discovering each other through the A2A Protocol backed by registry-enforced compatibility rules. You'll see agent cards registered via REST API, agents discover peers using the same patterns as OpenAPI discovery, and the registry reject incompatible version pushes. This brings proven API governance patterns to multi-agent AI systems. Key message for API audience: if you solved API discovery and versioning, you already know how to solve agent discovery. The patterns are identical.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4650,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "The API Called Back: When Untrusted Data Hijacks Autonomous AI Agents",
  "speakers": [
   {
    "name": "Kavya Pai",
    "title": "Associate Software Engineer",
    "company": "A.P Moller Maersk"
   }
  ],
  "facets": [
   "agent_readiness",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "error_semantics"
  ],
  "why": "Prompt injection through API responses hijacking agents.",
  "abstract": "The API Called Back: When untrusted data is used to control autonomous AI agents. A new threat has emerged when AI agents begin to interact with external APIs. Malicious prompts can be hidden within the API response that instruct the agent to call another API that shouldn’t be called. This talk will look into how we can use prompt injection to exploit trusted APIs. Why it may not be possible to prevent this with standard API security and how we can make more secure agent to API interactions by using least privilege, dynamic authorisation and validation. As autonomous agents become more common place in our business systems it is important we ensure that we are restricting what an AI agent can do with the information they receive.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4630,
  "day": "2026-12-02",
  "type": "Panel discussion",
  "start": "15:00",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Frictionless & Fully Secure: The Power of Mobile Network APIs",
  "speakers": [
   {
    "name": "Lina Taguengayte",
    "title": "Head of Innovation Projects",
    "company": "Bouygues Telecom"
   },
   {
    "name": "Dau Phi Nguyen",
    "title": "Value Added Services strategist",
    "company": "Free"
   },
   {
    "name": "Jill Cooper",
    "title": "Senior Go-to-Market Project Manager, Europe",
    "company": "GSMA"
   },
   {
    "name": "Guillaume Cayatte",
    "title": "Business Owner Telco APIs",
    "company": "Orange"
   },
   {
    "name": "Didier Cadars",
    "title": "Product Marketing Business Messaging and Digital Identity",
    "company": "SFR"
   }
  ],
  "facets": [
   "access_clarity",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity"
  ],
  "why": "No abstract; mobile network APIs for secure authentication, read from the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4588,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:05",
  "end": "15:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Cooperative Data Centers. Everyone wants them, nobody builds them. Why?",
  "speakers": [
   {
    "name": "Thomas Hutmacher",
    "title": "Co-Founder",
    "company": "collective zero e.V."
   },
   {
    "name": "Thomas Fricke",
    "title": "Member",
    "company": "CollectiveZero"
   },
   {
    "name": "Vincent Rahn",
    "title": "Co-Founder",
    "company": "collective zero e.V."
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Cooperative European data centers.",
  "abstract": "Currently, few global players dominate the data center market, mostly US-based, incompatible with EU law. With increasing demand for AI, they build massive facilities at the cost of water, energy and soil sealing. Switching to European providers helps, but will dependencies stay and only the name changes? What really matters is who owns the land and the building. Rarely, giant data centers are needed, and the ideal business model is a cooperative. Why doesn't it exist yet? We'll sketch a secure, eco-friendly cooperative data center, and invite for discussion: What does it take to get started?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4563,
  "day": "2026-12-02",
  "type": "Panel discussion",
  "start": "15:15",
  "end": "15:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "From Rack to Region: hands-on feedback on sustainable data centres and cloud",
  "speakers": [
   {
    "name": "Vincent Poncet",
    "title": "Cloud Sustainability Specialist",
    "company": "Google Cloud"
   },
   {
    "name": "Conor Molloy",
    "title": "Senior Projects Manager",
    "company": "Bureau Veritas"
   },
   {
    "name": "Emily Watson-Breeze",
    "title": "Purchaser Engagement Manager UK",
    "company": "TCO Development"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Cloud providers' GHG and efficiency data disclosure.",
  "abstract": "Companies have to report their GHG emissions and do due diligence of their sustainability credentials. Cloud providers deliver some data, both GHG emissions and efficiency metrics. I suggest a panel on those aspects with other providers and practitioners",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4542,
  "day": "2026-12-02",
  "type": "Panel discussion",
  "start": "15:15",
  "end": "15:55",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Vérité ou hallucination ?  Démystifions les idées reçues sur l’estimation de l’empreinte de l’IA !",
  "speakers": [
   {
    "name": "Aurore Tual",
    "title": "Digital Ecodesign Program Manager",
    "company": "Thales"
   },
   {
    "name": "Romain Bachelet",
    "title": "Référent Numérique Responsable",
    "company": "SNCF Voyageurs Matériel"
   },
   {
    "name": "Marine Le Touzé",
    "title": "Senior Manager Data & Partnerships",
    "company": "Schneider Electric"
   },
   {
    "name": "Rose Djagbre",
    "title": "Responsable Data Empreinte Carbone",
    "company": "Orange"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "A game on AI footprint estimation myths.",
  "abstract": "Un jeu à nul autre pareil! Des entreprises européennes de référence concourent pour les meilleures pratiques eco-responsables et tentent de gagner le plus grand arbre de tous les temps!",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4835,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Stéphane Requena",
    "title": "CTO",
    "company": "GENCI"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4657,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "From API Access to Authorized Action: Governing MCP Agents",
  "speakers": [
   {
    "name": "Shamlah Veerapandian",
    "title": "Lead Engineer",
    "company": "Pandora"
   }
  ],
  "facets": [
   "agent_readiness",
   "accountability",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [
   "delegated_identity",
   "consent_identity",
   "mcp_server"
  ],
  "why": "Who may use MCP tools, for which tasks, with what evidence.",
  "abstract": "Giving an engineering agent access to an API is easy. Deciding when it should be allowed to act is harder. MCP makes tools available to a model, but teams still need to define who can use them, for which tasks, and with what evidence. In this talk, I will explore that boundary using lessons from developing a software-factory prototype for enterprise engineering. We will follow a proposed workflow across tickets, repositories, tests and delivery systems, examining scoped permissions, evidence tied to a code revision, independent verification and human approval before consequential actions. A successful tool call does not tell us whether the resulting change should be merged or deployed. I will explain what local qualification and draft-PR CI have demonstrated, while distinguishing those results from proposed security controls and delivery integrations that remain unresolved. This is a practical discussion of a system still being developed, including its limits and open questions. Attendees will leave with a method for mapping agent-to-API trust boundaries, reviewing the evidence behind an action, and deciding where authorization should be enforced outside the model.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4651,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "The Human Side of Sovereignty: From Shadow AI to Cognitive Defense",
  "speakers": [
   {
    "name": "Isabelle Galy",
    "title": "Vice-Presidente",
    "company": "ClusterIA"
   }
  ],
  "facets": [
   "accountability",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Shadow AI and employee behaviour; organisational, not an API surface.",
  "abstract": "What if the real threat to your sovereignty wasn’t technology… but how your people use it? A new form of conflict is emerging : a cognitive war. Not fought with weapons, but through the tools we adopt every day. Employees are not resisting AI. They are embracing it, often through external, uncontrolled tools. What we call Shadow AI is not deviance. It’s adaptation. But behind this efficiency lies a shift: decisions influenced, knowledge shaped, dependencies created. This is not just an IT risk. It’s a strategic vulnerability. Because sovereignty today is not only about where your data is stored but who is shaping the way your people think. The real challenge is not to block these tools, but to transform your organization: From unaware users of Shadow AI to an army of vigilant defenders of your cognitive sovereignty. Because in a cognitive war, every employee is on the front line.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4570,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "From Cloud Billing to GreenOps Telemetry: Making Sustainability Observable",
  "speakers": [
   {
    "name": "Diana Todea",
    "title": "Head of Developer Relations Engineering",
    "company": "VictoriaMetrics"
   }
  ],
  "facets": [
   "outside",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "Billing exports turned into sustainability metrics; GreenOps telemetry.",
  "abstract": "Cloud sustainability data often lives apart from the systems engineers use every day. This session demonstrates a GreenOps pipeline that transforms Azure billing exports with SPRUCE, exposes enriched cost and sustainability data as Prometheus-compatible metrics in VictoriaMetrics, and visualizes them in Grafana. Whether in Europe or elsewhere, building sustainable and resilient infrastructure starts with giving engineers the tools to measure environmental impact and make informed decisions using open-source technologies.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4637,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "Models Reason, APIs Act: Defending the API Attack Surface in the Agentic Era",
  "speakers": [
   {
    "name": "Stephane Benfredj",
    "title": "Director API Security Architecture, EMEA",
    "company": "Akamai"
   }
  ],
  "facets": [
   "operational_transparency",
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "Shadow APIs, pre-launch testing and runtime threat detection for agent traffic.",
  "abstract": "AI agents accelerate API risks across sensitive data and tools. Learn to uncover shadow interfaces, test APIs before launch, and detect runtime threats as AI shifts from answering to taking action.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4734,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Event Contracts Are Domain Decisions",
  "speakers": [
   {
    "name": "Dr. Annegret Junker",
    "title": "Chief Software Architect",
    "company": "codecentric AG"
   }
  ],
  "facets": [
   "contract_quality",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "AsyncAPI drafts must carry domain decisions like broker binding.",
  "abstract": "Ask an LLM to draft an AsyncAPI document from a domain model and you get a valid file in seconds. In one of our projects, one such draft came back bound to Kafka. The team had decided on RabbitMQ months earlier. Nobody had told the model, and nothing in the domain model could have: a good domain model knows nothing about brokers. That small mistake points at a bigger question. An AsyncAPI document records decisions: which events exist, what they mean, who may rely on them, and how they evolve. AI can write the YAML. It cannot make those decisions. So where do they come from, and how do they reach the spec intact? This talk follows the event contracts of Larder, a cooking platform we modeled and built end to end in a project. We start where the events are found: an Event Model that marks which interactions are synchronous and which are asynchronous. We add what a board leaves open with an asynchronous API Product Canvas, a refined Visual Glossary that carries invariants, and an ADR that fixes the broker. Then AI drafts the AsyncAPI. Every operation description reads like a sentence from the Domain Story, so domain experts can review the contract without translation. We compare two generated event contracts side by side: one derived from the canvas, one derived straight from the aggregates. We carry SOLID over to event contracts — backward compatibility as Liskov substitution over time, additive evolution with tolerant readers, one AsyncAPI per Bounded Context. And we show what tooling catches and what it cannot: a diff finds a removed field, but only a check against the Ubiquitous Language finds a field that quietly changed its meaning. Finally, the contract becomes the spine of implementation: publishers and consumers come from established AsyncAPI code generators and are wired test-first with an outbox relay, and contract tests guard every release. The message is simple: the promise of a Bounded Context is made by people — in the canvas, the glossary and the decisions. AI writes the boring part: the AsyncAPI document that records it.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4326,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "The AI Agent Era: From Coding to Orchestration",
  "speakers": [
   {
    "name": "Akhara Sang",
    "title": "API/ DevOps Cloud Consultant",
    "company": "Naoplay"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "Shift from prompting to orchestrating agents; little API surface.",
  "abstract": "Until recently, working with AI followed a simple pattern: a human asks, the assistant answers. We prompt, validate, and prompt again. It's useful, but every step still depends on our attention. Something is starting to change. A new pattern is taking shape: human-led, agent-operated. The human sets the intent and the guardrails, then reviews the result. In between, the agent plans, acts, and iterates on its own. This talk looks at what that shift might mean in practice. How do we write instructions when the AI is going to act, not just answer? How do we build trust in an outcome we didn't watch unfold step by step? And where does this approach actually help, versus where it still falls short?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4502,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:30",
  "end": "15:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Amplifying APIs to Visuals for Cost Effective Business Analytics and Financial data",
  "speakers": [
   {
    "name": "Syed Khalid",
    "title": "Data Scientist",
    "company": "Adidas"
   },
   {
    "name": "Syed Usman Ahmad",
    "title": "Staff Developer Advocate",
    "company": "Grafana Labs"
   }
  ],
  "facets": [
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "Financial data APIs turned into visuals for analytics.",
  "abstract": "In the world of information, financial data plays a major role in decision-making for investors as well as companies. This data spans a wide array of sectors and financial instruments, including stock data, income statements, P&L statements, and so on. Although APIs are the backbone for providing the data foundation, the real challenge comes when they are not in a standardized format that adheres to your specific needs and requirements for your project. In the case of financial applications, this becomes even more critical/complex as the data spans a wide array of types; from numbers and figures to news and images. This raises the question of how we can effectively consolidate this into a single visual platform capable of providing a holistic solution to consumers while keeping it vendor-agnostic. In this talk, we will present a single unified market analytics dashboard using APIs with Grafana that can be leveraged by consumers and investors for data-driven decision making. This will be based on open-source tools that will provide an alternative to the costly solutions present in the industry while also avoiding vendor lock-in. After this session, the audience will have a better understanding of the possibilities and advantages that open-source tools provide in the financial sector and how they can be used with their APIs to create a scalable and robust financial research platform.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4560,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "15:45",
  "end": "16:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Community Sponsors' Pitching Time",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Sponsor pitches.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4407,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Stop Throwing Away Your Data - Towards Data-Driven API Governance",
  "speakers": [
   {
    "name": "Matthias Keckl",
    "title": "Software Engineer",
    "company": "qupaya technologies GmbH"
   }
  ],
  "facets": [
   "contract_governance",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "Feeding lint, test and latency data back into API governance.",
  "abstract": "A developer bypasses a linting rule to unblock a release, unknowingly shipping a breaking change. Meanwhile, a critical endpoint slows down release after release, never tripping a test timeout, until latency spikes in production. These problems reach production not because teams lack linting or testing tools, but because traditional pre-production checks operate as isolated pass-or-fail gates. The moment a check finishes, its rich diagnostic data vanishes into terminal outputs or CI logs. While developers focus on passing the current build, engineering leaders remain blind to accumulating technical debt until production monitoring alerts fire. But what if we stopped treating pre-production checks as throwaway gatekeepers and started treating them as continuous quality signals? In this session, we will explore what specific diagnostic data to collect across each stage of the API lifecycle. We will look at practical patterns for gathering this data without adding friction to daily developer workflows. Finally, we will outline how teams can use this raw execution data to build clear API health and governance signals, enabling leaders to spot trends and offer guidance long before code reaches production.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4394,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "Never Forward, Always Rebuild: Zero-Trust Payloads for APIs",
  "speakers": [
   {
    "name": "Marcel Grauwen",
    "title": "CPO",
    "company": "glomidco"
   }
  ],
  "facets": [
   "operational_transparency",
   "contract_quality",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "Rebuilding every payload at the trust boundary against its schema.",
  "abstract": "No Pass-Through: Rebuilding Every API Message at the Trust Boundary API gateways route traffic and WAFs block known-bad patterns, but what happens when an approved payload carries something nobody anticipated? Cross-domain guards solve this with a stricter rule: never forward a message as received. Parse it, inspect it, and rebuild it clean. This talk shows how UTL-X, an open-source, format-agnostic transformation language, becomes such a guard. Every incoming message, whether JSON, XML, CSV, YAML or OData, is parsed into one Universal Data Model, and its schema can be checked against an approved contract (XSD, JSON Schema and more). UTL-X 1.1's pure, deterministic validation layer then inspects that tree: field allow-lists, value constraints and cross-field checks. What passes is re-serialized into a strict canonical form, in the same format or a different one. Comments, duplicate keys, smuggled bytes and unexpected fields simply do not survive. What fails is rejected with an audit record. We'll walk through a concrete example: a partner's JSON order enters, only approved fields are kept, and canonical XML leaves for the internal system. You'll see why this \"\"parse, inspect, rebuild\"\" pattern matches NIST and NCSC cross-domain guidance, and how to apply it to your own APIs.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4562,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Green Software Landscape",
  "speakers": [
   {
    "name": "David Kopp",
    "title": "Green IT Consultant",
    "company": "envite consulting GmbH"
   }
  ],
  "facets": [
   "outside",
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "An open landscape of green software tools.",
  "abstract": "The green software tool ecosystem is growing fast, and it’s hard to know which tool fits which system. That’s why the \"Bundesverband Green Software\" (a German association) decided to create an open tool landscape: landscape.bundesverband-green-software.de As someone involved in creating and maintaining it, I’ll give an overview, show how to find the right tool for your needs (which metrics, which lifecycle phase, how accurate?), and make room to share your own tool experiences. We’ll also preview plans to extend the landscape beyond software, e.g. data centers and AI providers.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4330,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Your AI Agent Is Only as Good as Your API Description — Ensuring OpenAPI Quality at Enterprise Scale",
  "speakers": [
   {
    "name": "Pavel Kornev",
    "title": "Principal Software Architect",
    "company": "SAP SE"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "openapi_examples",
   "operation_distinctiveness"
  ],
  "why": "OpenAPI description quality as what agents actually read.",
  "abstract": "AI agents don't read documentation — they read metadata. As API-driven automation and agentic architectures become the norm, the quality of your OpenAPI descriptions is no longer a nice-to-have; it's the foundation that determines whether machines can reliably discover, understand, and orchestrate your services. Yet across the industry, most API descriptions remain incomplete, inconsistent, or outright misleading — and existing linting tools only scratch the surface. In this talk, we share how a large enterprise manages thousands of APIs while treating quality as a moving target. We'll have a look at the current validation tooling, expose its blind spots, and present the approach we built to enforce rich, machine-consumable API descriptions at scale — without slowing down the hundreds of teams that ship them. You'll learn how decentralized governance, automated validation pipelines, and a shift-left mindset can coexist with high developer velocity, and why getting this right is the single most impactful thing you can do to make your APIs AI-ready.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4836,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "Model Extraction: Why You Should Care and How to Protect Yourself?",
  "speakers": [
   {
    "name": "Marie Paindavoine",
    "title": "CEO",
    "company": "Skyld AI"
   }
  ],
  "facets": [
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; protecting models from extraction, read from the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4735,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "GraphQL as the Execution Layer for AI Agents",
  "speakers": [
   {
    "name": "Pascal Senn",
    "title": "Founder",
    "company": "ChilliCream"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "No abstract; GraphQL as the agent execution layer, read from the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4654,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Why Are We Still Governing AI Like Software?",
  "speakers": [
   {
    "name": "Dorian Blanc",
    "title": "Field CTO",
    "company": "Gravitee"
   }
  ],
  "facets": [
   "agent_readiness",
   "accountability"
  ],
  "agent_readiness_dimensions": [
   "agent_identity_declared",
   "delegated_identity"
  ],
  "why": "Agent identity, bounded authority and audit trail; vendor demo.",
  "abstract": "Agents don't reach production because trust is missing. Live with Gravitee: a governance layer giving an agent identity, bounded authority and an audit trail. Realistic scenario, public, replayable.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4569,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "From Green IT to Green IS: Lessons from a Central Bank",
  "speakers": [
   {
    "name": "Dr. Antonios Papargyris",
    "title": "IS Analyst",
    "company": "Bank of Greece"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Green IS at a central bank.",
  "abstract": "This talk presents BOGSIS (Bank of Greece’s Sustainable Information Systems), an initiative that moves beyond Green IT to a broader, sociotechnical Green IS perspective. Drawing on three case studies, combining SCI-based assessment with qualitative research, it explores sustainability across technology, usage practices and organizational processes. It shows how practitioners can make Information Systems more frugal by simplifying processes, reducing unnecessary data and building green digital skills, leading to 14 practical guidelines.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4631,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "I don't write code anymore: decisions behind an AI-Native, API-First Startup",
  "speakers": [
   {
    "name": "Fran Mendez",
    "title": "Founder",
    "company": "Commune"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "A one-person API-first startup built by an AI engineer.",
  "abstract": "Commune is a newsletter platform built and run by one person who no longer writes code. An AI engineer picks up tasks from Notion, builds them locally, and opens pull requests. I didn't design that setup upfront. It grew out of a string of decisions, some good and some I reversed the same day. This talk walks through them: Why the creator of an open source spec didn't go open source this time. What it's like to build on other platforms' APIs (Beehiiv, Kit, and Substack, which has none) when those platforms start competing with you. Why a paying customer, not a roadmap, decided when Commune shipped its own API and MCP server. How to design AI features users actually want instead of slop. You'll leave with the decisions, the mistakes, and what I'd do differently.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4546,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Building the essential, efficiently: Eco-Design redefines our Digital Services",
  "speakers": [
   {
    "name": "Laurent David",
    "title": "Green IT Manager",
    "company": "Worldline"
   },
   {
    "name": "Aurore Tual",
    "title": "Digital Ecodesign Program Manager",
    "company": "Thales"
   },
   {
    "name": "Christophe Hirotani",
    "title": "Référent Numérique Responsable CA-TS",
    "company": "Crédit Agricole Technologies et Services"
   },
   {
    "name": "Tala Awada",
    "title": "Sustainable IT Project Manager",
    "company": "VINCI Energie"
   }
  ],
  "facets": [
   "outside",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "ISO/IEC 20125-1 eco-design standard for digital services.",
  "abstract": "As digital usage booming across sector, how do we create high-value digital services while minimizing environmental impact? ISO/IEC 20125-1:2026, the first international eco-design standard for digital services (Feb 27, 2026), provides a clear lifecycle sustainability framework. In this conference you’ll discover: origins and implications of the standard; case studies from Worldline, Crédit Agricole, Thales, VINCI Energie; how to embed eco-design in your product; future standards. For developers, PMs, service managers, CIOs. Build more value, not just more.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4413,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:30",
  "end": "16:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Taming the API & SDK Zoo",
  "speakers": [
   {
    "name": "Matthieu Croissant",
    "title": "Principal Scientific Product Leader",
    "company": "Roche AG"
   }
  ],
  "facets": [
   "contract_governance",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "Governance, DX and SDKs across Roche R&D integrations.",
  "abstract": "Bringing order into your API landscape takes more than wishes, we will talk about our journey as the R&D Unit of Roche to bring governance, better DX, and faster delivery to our latest integrations and scientific workflows.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4535,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:55",
  "end": "17:20",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Les audits d'écoconception ne servent à rien (la plupart du temps)",
  "speakers": [
   {
    "name": "Laurent Devernay Satyagraha",
    "title": "Consultant",
    "company": "LauDevSat"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Why eco-design audits go nowhere.",
  "abstract": "J'ai effectué, lu et relu des tonnes d'audits d'écoconception. Vous savez comment ça se passe : ce n'est pas simple d'effectuer un audit (au moins au début). À la fin, vous présentez les résultats et... rien ne se passe. Je vous donnerai ici des astuces pour capter l'attention et avoir un effet pérenne. Aussi, nous verrons comment considérer l'écoconception comme un tout afin d'aller au-delà des audits, rustines et autres solutions miracles.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4558,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "16:55",
  "end": "17:20",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Demonstrating the Real Value of Digital Sobriety through ROI-MPACT",
  "speakers": [
   {
    "name": "Angelica Calvet",
    "title": "Vice-président",
    "company": "Alliance Green IT"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "ROI of digital sobriety.",
  "abstract": "This talk presents ROI-MPACT Digital Sobriety, a methodology developed within the Alliance Green IT (AGIT) to assess the ROI of digital sobriety initiatives through a multi-capital approach. Beyond traditional financial ROI, it values avoided energy use and emissions, deferred CAPEX, reduced risks, freed-up human time and improved service quality, helping IT, finance, CSR and business teams turn responsible digital initiatives into robust, measurable and sustainable business cases",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4576,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "16:55",
  "end": "17:20",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "What the Water Sector Teaches Us About Making Green Tech Actually Stick",
  "speakers": [
   {
    "name": "Mallak Bani Mustafa",
    "title": "Innovation Project Manager",
    "company": "Southern water"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Green tech adoption lessons from the water sector.",
  "abstract": "I’ll share lessons from working in the water and utilities sector, where sustainability, climate adaptation, and infrastructure constraints collide daily. The talk focuses on why many GreenTech solutions struggle to scale beyond pilots and what the water sector has learned about designing, deploying, and governing technology for long‑term resilience and real impact.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4360,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Agents Inherit Your Governance — Whether You Meant Them To or Not",
  "speakers": [
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance",
   "contract_quality",
   "discoverability"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "mcp_server",
   "auth_clarity"
  ],
  "why": "The Kin Score and Agent Readiness themselves: agents inherit governance.",
  "abstract": "Every API provider in the APIs.io catalog, roughly 25,000 companies, carries two scores derived from years of API Evangelist profiling. The \"Kin Score\" is a 0–100 composite of governance fundamentals: contracts, documentation, security, support, operations, and a conditional regulatory layer that applies only to banking, health, payments, insurance, securities, and government. Agent Readiness is a separate 0–100 rating across twelve dimensions, measuring whether an autonomous agent can drive the API safely without a human filling in the gaps. The two are scored independently and never blended. That was a deliberate design choice, and it turned out to be the most useful one, because it lets the catalog answer a question the industry keeps asserting without evidence: does governance actually determine AI readiness, or is agent readiness something you can ship separately? The data shows that you inherit it, Agent readiness is a layer, and a layer cannot rise above its foundation. The dimensions that separate the agent-native band from everyone else, idempotency, error semantics an agent can branch on, auth clarity, consent and identity, rate-limit signalling, are not AI features. They are governance disciplines that were always supposed to be there, now load-bearing in a way they weren't when a human developer was around to work around them. Meanwhile a very large cluster of providers sits at an identical baseline: a spec, an auth scheme, a rate-limit header. That is the floor of the enrichment pipeline, not readiness. My talk will walk the composite rubric facet by facet, shows what each one becomes once an agent is the consumer, and then reads the scored catalog for what actually moves the agentic readiness layer. It includes the negative finding: shipping an MCP server on top of an ungoverned API does not move the number, and in a governed portfolio it makes things worse by giving agents a confident interface to an unreliable contract. As API governance moves from an offensive strategy to a defensive one in a market being flooded with agentic activity, the API foundation you are building on will matter when it comes to adapting to the unpredictable and chaotic market we are all trying to make sense of today.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4943,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "You Cannot Bolt Agent Readiness Onto Ungoverned APIs",
  "speakers": [
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "spec_presence",
   "auth_clarity",
   "error_semantics"
  ],
  "why": "Agent readiness cannot be bolted on; the two API Evangelist instruments.",
  "abstract": "\"\"AI-ready\"\" is now claimed by every API program and measured by almost none. API Evangelist scores roughly 25,000 providers on two separate instruments: the Kin Score, a 0–100 composite of governance fundamentals with a conditional regulatory layer for regulated industries, and Agent Readiness, a standalone 0–100 rating across twelve dimensions. They are computed independently and never blended — which means the relationship between them is a finding, not an assumption. The finding is that agent readiness is a layer built on the governance foundation, and it does not rise above it. An agent does not consume an API; it tries to complete a business capability — open the account, file the claim, move the shipment, settle the payment — and it fails at whichever contract in that chain is ungoverned. Readiness is therefore a property of a capability path, not of an endpoint, and the weakest governed link sets the ceiling for the whole path. This is why organisations that map their capabilities first and score their contracts second get a usable answer, and why organisations that ship an MCP server first get a confident interface to an unreliable contract. My talk works down from business capability to capability path to individual contract, showing which governance facets carry the weight at each level, and what the scored catalog says about how far most API portfolios actually are. As API governance moves from an offensive strategy to a defensive one in a market being flooded with agentic activity, the API foundation you are building on will matter when it comes to adapting to the unpredictable and chaotic market we are all trying to make sense of today.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4367,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "API Security: From Fundamentals to the Agentic Era",
  "speakers": [
   {
    "name": "Jesus Vallejo Garcia",
    "title": "Head of APIs",
    "company": "cloudappi"
   }
  ],
  "facets": [
   "operational_transparency",
   "regulatory",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity"
  ],
  "why": "API security across the lifecycle into the agentic era.",
  "abstract": "In this talk, we will cover the foundations of API security, starting from the very beginning at the definition stage. We will review how to integrate security throughout the lifecycle, the value of pentesting, and the critical role of tools like WAFs and API Managers. To wrap up, we will analyze the shift into the agentic era, exploring how the risk landscape changes when API consumers are no longer just humans, but autonomous AI agents.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4417,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Anthropomorphism is an API: designing the human interface of AI agents",
  "speakers": [
   {
    "name": "Stanislas Bernard",
    "title": "CTO of the AI Coworker offer",
    "company": "Theodo"
   }
  ],
  "facets": [
   "accountability",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "agent_identity_declared"
  ],
  "why": "Specifying an agent's human interface: name, tone, disclosure.",
  "abstract": "Every API comes with a contract: what it does, what it won't do, and how it fails. AI agents now have a second interface that nobody specifies: the human one. Their name, their tone, how they sign an email, when they take initiative, and whether they admit they are not human. For the past months, we have run AI chiefs of staff for executives at Theodo. They read mail, prepare meetings, draft replies, and post on Slack. Our users gave them names and faces, and delegation went up. Then the people on the other side of the email started noticing, and some felt fooled. Why does it matter? Because agents are leaving the chat window. They now act in the channels where humans work, in front of people who never chose to talk to a machine. In that world, anthropomorphism is not a UX gimmick. It is an interface, and an underspecified one breaks trust the way an undocumented breaking change breaks clients. The EU AI Act already makes part of that contract mandatory. In this talk, I treat the human side of an agent as an API: what to specify, what to version, and what to never fake. You will leave with a concrete checklist to design agent personas that people delegate to without feeling deceived.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4340,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Green Architecture: less fat, more impact, more efficiency!",
  "speakers": [
   {
    "name": "Thierno Diallo",
    "title": "Staff Engineer",
    "company": "AXA France"
   }
  ],
  "facets": [
   "outside",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [],
  "why": "API Green Score applied to an API; payload efficiency is outside the rubric.",
  "abstract": "Have your APIs become a bit bloated? Are they consuming more resources than necessary? Don't panic, we're bringing out the toolbox to give them a green makeover! In this session, we'll automatically apply the \"API Green Score\" framework to a real, existing API, using open-source tools, concrete metrics, and a report that everyone can use. The observation is simple: modern APIs are often resource-intensive. The problem? Not a lack of principles, but the absence of measurable and automated rules. In this live session, we will: take measurements (payload, latency, cache), calculate the Green Score per endpoint, implement actionable rules: GZIP/Brotli compression, partial fields, streamlining of transferred data, mandatory pagination, etc., automate regression testing (SonarQube, Spectral). A dashboard to make results accessible The result? A lighter, faster API and a reduced environmental impact. We'll show you the gains live (before/after): reduced network traffic, stabilized latency, and optimized caching. And to take it with you, a GitHub repository awaits you with everything you need to replicate the process.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4743,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "AsyncAPI in the Agentic Era: Giving AI Agents Ears",
  "speakers": [
   {
    "name": "Laurent Broudoux",
    "title": "Co-founder",
    "company": "Microcks & Reshapr"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality",
   "discoverability"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described",
   "spec_presence"
  ],
  "why": "AsyncAPI contracts describe events agents can listen to and trigger.",
  "abstract": "AI agents can call your APIs, but they are blind to what happens in your system. Meanwhile, your Kafka topics, message queues, and AsyncAPI contracts already provide a precise description of everything that happens and everything that can be triggered. So, how do you connect the two? This session explores the role of AsyncAPI within agentic workflows. We will begin by addressing a specific limitation: the request/response model of the Model Context Protocol (MCP) forces an agent to rely on polling, whereas event-driven architectures are asynchronous by nature. We will examine ongoing developments within the MCP specification—such as subscriptions and notifications—and explore how an AsyncAPI contract can support them. A live demo with the open source reShapr project and Kafka shows it working. We’ll also cover the open questions: delivery semantics, filtering, auditing, tracing and access control. You’ll leave with a practical lens on which events to expose to agents, and what your AsyncAPI contract needs to become an agent-ready on.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4837,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "Specialists Aren’t Born. They’re Fine-Tuned.",
  "speakers": [
   {
    "name": "Mohamad Alhajar",
    "title": "CEO",
    "company": "legml.ai"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; title is about fine-tuning models, not APIs.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4636,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:00",
  "end": "17:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Agents Need APIs: Multi-Agent Pipelines in Production",
  "speakers": [
   {
    "name": "Camille Epitalon",
    "title": "CTO & Co-founder",
    "company": "Verso"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "mcp_server"
  ],
  "why": "APIs designed for AI callers first, exposed through UI, CLI and MCP.",
  "abstract": "We run Verso, an agentic platform for qualitative consumer research: voice AI interviews at scale, multi-agent pipelines, and a company brain that agents and humans share. Clients hit us through a UI, a CLI, and MCP — so our APIs are designed for AI callers first, not only humans. This talk is a production postmortem. I’ll show how we harness multi-agent systems that run thousands of hours of interviews a day: tool contracts, sandboxing, evals, and the failure modes that only show up at volume. I’ll cover what changes when the primary consumer of your API is an agent (idempotency, schemas, observability, rate limits), how MCP and CLI become the real product surface, and how a versioned company brain keeps tacit knowledge in the loop instead of trapped in Slack. You’ll leave with a concrete checklist for shipping agentic systems that talk to real APIs — less demo magic, more boring reliability. Based on building Verso (seed-backed, enterprise clients) as both an agentic product and a semi-autonomous company run with agents.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4544,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:20",
  "end": "17:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Penser un e-commerce durable à l’ère de l’IA : usages et impacts",
  "speakers": [
   {
    "name": "Perrine Tanguy",
    "title": "Fondatrice",
    "company": "Collectif E-CO, E-Commerce Responsable"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI's environmental impact on e-commerce; no facet reads sustainability.",
  "abstract": "Fondée sur l’étude 2026 d’E-CO, cette conférence explore comment l’IA transforme les pratiques du e-commerce et leurs impacts environnementaux, sociaux et éthiques. Elle met en lumière les risques clés, de la consommation d’énergie au biais d’automatisation, et identifie des leviers pour un usage plus responsable. À partir de cas concrets, elle introduit le concept d’« IA robuste » et propose des principes actionnables pour concilier performance, innovation et durabilité.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4934,
  "day": "2026-12-02",
  "type": "Workshop",
  "start": "17:20",
  "end": "17:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Et si votre IA apprenait la sobriété ?",
  "speakers": [
   {
    "name": "Guillaume Gallon",
    "title": "Architecte IT et Numérique Responsable",
    "company": "Institut du Numérique Responsable"
   }
  ],
  "facets": [
   "outside",
   "open_source"
  ],
  "agent_readiness_dimensions": [
   "agent_skills"
  ],
  "why": "Open-source sobriety skills for coding assistants; environmental frugality is unmeasured.",
  "abstract": "L'IA peut-elle contribuer à un numérique plus responsable alors qu'elle en est elle-même l'un des défis environnementaux ? À l'INR, nous avons choisi d'explorer cette voie avec des skills open source pour Claude Code, Codex, Mistral Vibe et 8 autres assistants IA afin d'intégrer les bonnes pratiques du numérique responsable (écoconception, accessibilité, sobriété…) directement dans le code produit par l'IA, tout en limitant sa propre consommation de ressources et de tokens. Des outils gratuits à découvrir, tester et enrichir ensemble.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4585,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:20",
  "end": "17:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Green IT, Smart Business: How City of Copenhagen Cuts Carbon and Costs at the Same Time",
  "speakers": [
   {
    "name": "Kerstin Roberts",
    "title": "IT Category Manager and Sustainability Lead",
    "company": "City of Copenhagen"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Green IT procurement and cost at a city; no API surface.",
  "abstract": "At the City of Copenhagen, Green IT is not a cost – it’s a shortcut to efficiency. By rethinking how we buy, run and retire technology, we cut energy use, extend the life of our equipment and reduce waste, while lowering our total IT spend. We use data on both CO₂ and costs to steer decisions, so sustainable choices become the smartest business choices. In Copenhagen, Green IT is a strategic tool: it powers better digital services, protects the climate and strengthens the bottom line at the same time.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4393,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "From Friction to Platform: Rebuilding a Permission System at Scale",
  "speakers": [
   {
    "name": "Elizabeth Zaslavsky",
    "title": "Lead Product Manager - Fintech API Platform",
    "company": "Ippon"
   }
  ],
  "facets": [
   "access_clarity",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "auth_clarity"
  ],
  "why": "Rebuilding fragmented permissions as an OpenFGA ReBAC platform.",
  "abstract": "From a fragmented permission system (7 permission types, 17 endpoints, up to 3 weeks lead time) to a ReBAC platform built on OpenFGA — a progressive rewrite, no big bang, onboarding around thirty products and teams. When tech strategy serves a product vision and real business stakes.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4632,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "The API Multiverse: One Internet, Infinite Agents, New Rules",
  "speakers": [
   {
    "name": "Aura Yavary",
    "title": "Researcher, Founder and Innovator",
    "company": "UC Davis"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability"
  ],
  "agent_readiness_dimensions": [
   "well_known_catalog",
   "agent_identity_declared"
  ],
  "why": "Agents discovering and using APIs as primary consumers.",
  "abstract": "The internet was designed for humans to navigate and APIs for applications to communicate. But what happens when autonomous AI agents become the primary users of both? We are entering the API Multiverse: a world where thousands of intelligent agents discover services, negotiate capabilities, orchestrate workflows, and interact across organizational boundaries. Yet today's API architectures were largely designed around predictable requests, predefined integrations, and human-controlled execution—not autonomous systems that reason, make mistakes, adapt, and act. This talk explores the technical foundations of an agent-native internet and proposes a practical architecture for building it. We will examine how API contracts must evolve to expose machine-readable capabilities, how agents can coordinate through dynamic discovery and orchestration, and why identity, authorization, observability, and verification become essential when AI systems initiate real-world actions. Drawing on research in multimodal world models, autonomous decision-making, and failure-aware agents, the session will introduce an actionable framework for reliable agent–API interactions, including state tracking, failure detection, recovery mechanisms, and human oversight. Attendees will leave with concrete architectural patterns, design trade-offs, and a roadmap for adapting existing API ecosystems to autonomous agents. The central question is no longer how to connect applications. It is how to build an internet where intelligent agents can act independently without sacrificing reliability, security, or human control.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4408,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "From APIs to Outcomes: Building Ecosystems Around Human Needs",
  "speakers": [
   {
    "name": "Heather Xiao",
    "title": "Founder and CEO",
    "company": "Horizon Zero"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Ecosystem outcomes for people; no published-surface signal.",
  "abstract": "APIs connect systems, organisations and services. But how do those connections translate into better outcomes for people, and sustainable value for the organisations involved? Consider a household facing a sudden loss of income. Its bank sees falling deposits, its energy supplier sees missed payments, and its local authority may hold information about available support. Each organisation sees part of the picture. How could APIs enable them to identify pressures earlier and coordinate appropriate help, with the person’s permission? This talk introduces the Human Needs-to-Outcomes framework, originated through Horizon Zero and being explored in collaboration with Warwick Business School. It offers a practical approach to ecosystem design: start with human needs and desired outcomes, then work backwards to the services, intelligence, data and API capabilities required. Using VIDI/GlowTrack as an illustrative application, the session explores how this approach informs decisions about cross-sector orchestration, consent, interoperability and accountability. It also examines the economics of participation: who contributes, who benefits, who pays, and why organisations would collaborate. For API leaders, architects and ecosystem builders, the challenge is to turn connectivity into services people use and organisations will sustain. Attendees will leave with a way to connect API strategy to human outcomes, business incentives and measurable impact.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4369,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6/7"
  ],
  "title": "60,000 CVEs and Counting: What 60 Years of AppSec Tell Us",
  "speakers": [
   {
    "name": "Pedram Hayati",
    "title": "CEO",
    "company": "SecDim"
   }
  ],
  "facets": [
   "operational_transparency",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "AppSec history and CVE volume; the security disclosure posture.",
  "abstract": "2026 passed 60,000 published CVEs before the end of September, more than any previous full year and roughly 50% up on 2025. But raw vulnerability volume is only part of the story. Application Security has evolved far beyond just fixing injection flaws. Over the past two decades, it has matured into a complex ecosystem encompassing automated code scanning, secure CI/CD pipelines, cloud-native security posture management, and most recently, AI copilots and autonomous remediation. In this session, we will walk through the key milestones that have shaped modern AppSec. We will unpack emerging threats, the rise of AISecOps, the growing mandate for SBOMs, and the dual-edged role of AI in both attacking and defending software. Why does this matter? Vulnerability volume is growing faster than ever, and AI is now both writing and fixing code. Understanding how AppSec got here is essential to deciding what comes next. You will leave with a clear understanding of today's AppSec landscape, the challenges that lie ahead, and the strategies needed to secure applications in 2026 and beyond.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4358,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "From API Sprawl to API Clarity",
  "speakers": [
   {
    "name": "Meryem Benaich",
    "title": "Founder (ex Data Manager/ CDO)",
    "company": "Clarus Data Management"
   }
  ],
  "facets": [
   "contract_governance",
   "discoverability",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "API sprawl, undocumented endpoints and governance in regulated settings.",
  "abstract": "APIs have become the backbone of modern digital ecosystems. Yet in many organizations, especially in industrial and regulated environments, API growth outpaces governance. What begins as acceleration quickly turns into fragmentation: undocumented endpoints, unclear ownership, inconsistent standards and rising compliance exposure. This talk explores a pattern observed across multiple mid-market transformation programs: API sprawl is rarely a technical failure. It is a visibility and prioritization failure. When APIs scale faster than control mechanisms, organizations lose traceability, weaken accountability and increase operational risk. Drawing from field experience in manufacturing and regulated industries, I will share a practical framework to assess API maturity, identify hidden risk layers and reintroduce structural clarity. Instead of treating governance as bureaucracy, we will examine how it becomes a strategic accelerator — enabling trust, interoperability and AI readiness. APIs are no longer just integration tools. They are control surfaces that shape how organizations scale, collaborate and remain compliant. Designing them with clarity is no longer optional — it is a condition for sustainable growth.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4838,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "GenerationAI"
  ],
  "room": [
   "Acacia 3"
  ],
  "title": "How AI Agents work?",
  "speakers": [
   {
    "name": "Emmanuel Orozco",
    "title": "DevRel",
    "company": "Leroy Merlin"
   }
  ],
  "facets": [
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "Explainer of the agent loop; no specific surface.",
  "abstract": "AI agents can write code, run tests, and automate workflows. But what’s actually happening under the hood? Surprisingly, it starts with a while loop. Through step-by-step examples, we’ll see how it chooses an action, executes it, reads the result, and decides what to do next. You’ll leave with a practical mental model of how agents work, and how to move from using an agent as a black box to understanding, and building, one yourself.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4883,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "OpenAPI: The Answer Key for MCP Tool Calling",
  "speakers": [
   {
    "name": "Miguel Quintero",
    "title": "Technical Trainer",
    "company": "Postman"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality",
   "spec_presence",
   "operation_distinctiveness"
  ],
  "why": "OpenAPI tool names, descriptions and schemas drive MCP tool selection.",
  "abstract": "Your MCP server passes its unit tests. Then an agent connects, picks the wrong tool, sends a well-formed date for the wrong week, and reports success. Failures like these come from how a model reads your tool names, descriptions, and schemas, and most teams have no way to measure them before release. This talk argues that the answer key already exists: your OpenAPI description. The same document that generates your MCP tools can grade how agents use them. Examples become golden test cases. JSON Schema becomes argument assertions. HTTP semantics and deprecation metadata become safety gates. We'll walk through a pipeline that lints the spec for agent readiness, generates the tools, scores tool selection, arguments, and task completion, and blocks a release in CI when an unsafe call slips through. You'll leave with a mapping from OpenAPI fields to eval signals you can apply to your own specs, a pipeline that fits into existing CI, and a case for holding destructive and permission-boundary calls to a zero-tolerance threshold instead of a blended pass rate.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4729,
  "day": "2026-12-02",
  "type": "Talk",
  "start": "17:30",
  "end": "17:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Best Practices Building Agentic AI GTM",
  "speakers": [
   {
    "name": "Andrey Lokot",
    "title": "Founding Commercial Lead",
    "company": "DataWave"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Selling agentic AI into banks; go-to-market, not API surface.",
  "abstract": "Everyone is racing to ship agentic AI, but almost nobody talks about what it actually takes to sell it, especially into regulated, risk-averse enterprises like banks. At DataWave, I've spent the last two years building and selling agentic AI systems into financial institutions, and most of the standard AI GTM playbook falls apart on contact with real buyers: procurement teams that don't trust automation, compliance officers who need to explain every decision an agent makes, and champions who need to defend the purchase internally long after the demo ends.This talk shares the practical, sometimes counterintuitive lessons from doing this in production: why we deliberately pushed complexity to the front end instead of hiding it behind a slick agent interface, why full-automation pitches kill deals faster than they close them, and how reframing the sale around augmentation rather than autonomy changed our win rate. Attendees will leave with concrete GTM tactics for positioning, pricing, and pitching agentic AI products to enterprise and regulated buyers, grounded in real deals rather than theory.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4568,
  "day": "2026-12-03",
  "type": "Opening remarks",
  "start": "09:15",
  "end": "09:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Opening remarks · The Detour Tax",
  "speakers": [
   {
    "name": "Gael Duez",
    "title": "Founder & Host",
    "company": "Green IO"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; opening remarks.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4639,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Microsoft talk",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; sponsor talk with no stated topic.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4858,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Lean AI: How Airbus Turns the Tide on GenAI Hype",
  "speakers": [
   {
    "name": "Marcel Mateescu",
    "title": "Business Relationship Manager - Generative AI",
    "company": "AIRBUS"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; GenAI strategy at Airbus, no API signal in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4857,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Can Harness Engineering make Smaller AI Models Viable?",
  "speakers": [
   {
    "name": "Naresh Jain",
    "title": "Founder & CEO",
    "company": "Specmatic"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "A behaviour harness for AI-driven API development with smaller models.",
  "abstract": "As coding agents become more capable, the default response to difficult engineering tasks is often to reach for a more powerful model. But what if we could improve the environment around the model instead? I built a behaviour harness for AI-driven API development using OpenAPI and AsyncAPI specifications, representative examples, contract testing, service virtualisation, API coverage, and deterministic feedback. The idea was simple: give a coding agent better guidance and faster feedback, and it should complete the task with fewer tokens and at lower cost. The experiment produced the opposite result. With a frontier coding model, the harness made the task nearly five times more expensive and twice as slow, with little difference in the final outcome. The model was already capable enough to solve the problem, and the harness became overhead. But when I repeated the experiment with a substantially smaller model, the result changed. Without the harness, the model could not successfully complete the task. With the harness, it could, at approximately one-eighth the token cost of the unharnessed frontier model, although with some reduction in code quality. This talk walks through the experiment, including the API development task, behaviour harness, isolated test environments, evaluation criteria, results, and limitations. More importantly, it explores a different way of thinking about AI-assisted software development: perhaps the purpose of a good harness isn’t to make an expensive model cheaper, but to make an expensive model unnecessary.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4494,
  "day": "2026-12-03",
  "type": "Keynote",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "What Europe's Government APIs Look Like From the Outside-In",
  "speakers": [
   {
    "name": "Kin Lane",
    "title": "Founder",
    "company": "API Evangelist"
   }
  ],
  "facets": [
   "discoverability",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; government APIs read from outside, which is how the score reads them.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4737,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Beyond the Query: GraphQL Features That Make Frontends Simpler",
  "speakers": [
   {
    "name": "Aurélien David",
    "title": "Engineering Manager",
    "company": "Pennylane"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; GraphQL features for frontend consumers.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4660,
  "day": "2026-12-03",
  "type": "Keynote",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "TBA",
  "speakers": [
   {
    "name": "Rory Blundell",
    "title": "CEO",
    "company": "Gravitee"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; TBA.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4669,
  "day": "2026-12-03",
  "type": "Opening remarks",
  "start": "09:15",
  "end": "09:40",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Opening remarks: who verifies the agents?",
  "speakers": [
   {
    "name": "Kapil Bakshi",
    "title": "CEO/Co-founder",
    "company": "Karate Labs"
   }
  ],
  "facets": [
   "agent_readiness",
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; opening remarks on who verifies agents.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4575,
  "day": "2026-12-03",
  "type": "Keynote",
  "start": "09:30",
  "end": "10:05",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Salt, Fat, Sugar, Data",
  "speakers": [
   {
    "name": "Tom Greenwood",
    "title": "Co-founder",
    "company": "Wholegrain Digital"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Food systems and data addiction; no API relevance.",
  "abstract": "For the first time in history, humans now live in a food system designed not for health or sustainability, but for addiction and profit. Fast food is engineered to stimulate a dopamine response through the constant drip of refined salt, sugars, and fats, stimulating our cravings for instant pleasure and contributing to a myriad of chronic health conditions. This pattern is mirrored in our digital world that now erodes our mental and societal wellbeing. This talk will explore how the industrialisation of food is now mirrored in the industrialisation of information, and what we can do about it.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4362,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "APIs as the Nervous System: Building a Trusted, Resilient Agent Fleet on Your Own Infrastructure",
  "speakers": [
   {
    "name": "Tuhin Mitra",
    "title": "Engineering Manager",
    "company": "Doctolib"
   }
  ],
  "facets": [
   "agent_readiness",
   "accountability",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [
   "delegated_identity",
   "auth_clarity"
  ],
  "why": "Agent fleet wired through APIs with audit trails and access control.",
  "abstract": "\"Every engineering team has a Slack channel where requests go to die. Bug reports without repro steps, doc questions that take hours to answer, access requests that bypass the audit trail. We built a fleet of agents that resolves them autonomously — wired entirely through APIs, designed to degrade gracefully, and running on infrastructure we control. The system classifies every incoming message and routes it to a specialised pipeline. Documentation questions trigger an agentic retrieval loop that calls the GitHub Code Search API, Confluence CQL, and curated FAQ endpoints iteratively — no vector index, no embeddings, just API calls and reasoning. Bug and feature reports pass an actionability gate before dispatching to a code exploration pipeline that calls the JIRA REST API to create a ticket and the GitHub REST API to open a draft PR. Resilience is built into every API boundary. Each pipeline node fails independently — a failed GitHub PR call doesn't roll back the JIRA ticket or silence the Slack reply. An offload mode dispatches to GitHub Actions via workflow_dispatch, so agent commits are server-signed and pass protected branch policies. PII is stripped before any external API receives user data. Prompt injection is neutralised before it reaches the LLM. This is a direct answer to Black Box Dependency: the brains of this system call your APIs, run on your infrastructure, and operate under your rules. The talk covers the API design decisions that made it production-ready, the failure modes we didn't anticipate, and what sovereign agentic architecture looks like in practice.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4406,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "The Cognitive Firm: Designing Human Control into AI Workflows",
  "speakers": [
   {
    "name": "Jorge Fábrega",
    "title": "Associate Professor at Universidad del Desarrollo (CHILE) and Partner at Datavoz",
    "company": "Universidad del Desarrollo & Datavoz"
   }
  ],
  "facets": [
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Designing human control and accountability into AI workflows.",
  "abstract": "Delegation to AI systems reduces the effort devoted to producing an output, yet it raises the demands placed on problem formulation, task decomposition, evaluation, integration, and accountability. The central organizational question is therefore how cognitive work is redistributed when AI enters a workflow, and whether humans retain the capacity to understand, assess, and justify the resulting output. In this talk, I develop the idea of The Cognitive Firm as an organizational architecture for allocating cognitive operations between humans and AI. Its central concern is Epistemic Ownership: the extent to which people remain able to direct a task, recognize the basis of its results, and incorporate those results into decisions for which they remain responsible. Evidence from the Anthropic Economic Index suggests that this redistribution differs across modes of use. API-based use is associated with human direction being located further upstream, where objectives, constraints, and validation rules are embedded in technical design. Conversational systems retain more direction within an iterative exchange, where users refine, evaluate, and accumulate understanding during execution. These patterns suggest that automation does not simply remove human cognitive work. It changes where that work occurs and which capabilities become organizationally decisive. My talk translates this trade-off into principles for designing AI workflows that increase productive delegation without hollowing out human judgment.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4640,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Agents Are Not Missing Tools. They Are Missing Business Context",
  "speakers": [
   {
    "name": "Kseniia Ilchenko",
    "title": "Lead Product Manager AI",
    "company": "DataGalaxy"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality"
  ],
  "why": "MCP tools with one-line descriptions lack the business context agents need.",
  "abstract": "For years, we’ve talked about data governance and decisions based on trusted data. Then agentic AI arrived, and we started giving agents access to MCP tools with a one-line description, expecting them to figure out how our business works. An API spec tells an agent what to call. It doesn’t tell it what “active user” means, who owns that definition, or who is allowed to access the data. Consider an agent asked to identify inactive customers for a retention campaign. Does “inactive” mean no login in 30 days, no purchase in 90 days, or an expired subscription? Should enterprise accounts be treated differently? Every tool works as expected, and the agent still produces the wrong list and triggers the wrong campaign. Humans know these things because they understand the business and its unwritten rules. Agents don’t. They guess. And when an agent gets it wrong, it may have already taken action before anyone notices. Adding an agent isn’t the hard part. Giving it the right business context is. That means making definitions, data lineage, ownership, and access rules explicit, and ensuring they’re enforced by the systems agents use rather than left for the model to infer. Drawing on our experience building an MCP server that provides governed business context to agents, I’ll walk through how we approached it: choosing which tools an agent should have access to, grounding its decisions in reliable context, and deciding when it has enough information to act - and when it should ask instead of guess.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4663,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Building a World Model for a $14 Trillion Industry: Why LLMs Alone Are Not Enough",
  "speakers": [
   {
    "name": "Saurabh Mishra",
    "title": "Founder and CEO, Professor",
    "company": "Taiyo.AI, SciencesPo, OECD.AI"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Industry world model beyond LLMs; no API surface.",
  "abstract": "What would it take to build an AI system that actually understands an industry rather than merely generating text about it? Infrastructure and construction provide an especially demanding test. Projects extend across decades, cross national borders and involve interdependent contractors, financiers, suppliers, locations, regulations and risk events. The relevant knowledge is fragmented, constantly changing and often contradictory. At Taiyo.AI, we have built the intelligence layer for this physical economy using millions of project records, large-scale procurement intelligence, entity relationships, geospatial signals and specialized AI agents. This session explores the architecture and lessons behind that effort: Context: How to organize heterogeneous data into domain-specific representations with temporal, spatial and entity relationships. Agents: How to orchestrate specialized reasoning workflows across opportunity discovery, project analysis, procurement and risk. Harness: How evidence provenance, citations, uncertainty, tool interfaces and evaluations make agents more reliable. Loops: How feedback, institutional knowledge and continuous data updates improve the system over time. The underlying thesis is that frontier models are a necessary component of enterprise AI, but they are not a substitute for the structured intelligence infrastructure needed to make consequential decisions. The discussion will also explore architectural limitations, trade-offs between models and data, and the emerging opportunity for domain-specific world models across the physical economy.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4670,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Karate Agent: what an agent can actually verify, and what it cannot",
  "speakers": [
   {
    "name": "Peter Thomas",
    "title": "Co-Founder & CTO",
    "company": "Karate Labs"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; what a testing agent can verify.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4351,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "From Pilot to Production: The Enterprise AI Enablement Stack",
  "speakers": [
   {
    "name": "Frank Kilcommins",
    "title": "Head of Enterprise Architecture",
    "company": "Jentic"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality",
   "well_known_catalog"
  ],
  "why": "Making enterprise APIs AI-ready: surface, describe and dynamically discover tools.",
  "abstract": "Scaling AI is an enablement problem, not just a model problem. This session introduces five foundational steps to make enterprise APIs AI-ready: (1) surface more tools/APIs systematically, (2) measure and describe them better, (3) enable dynamic discovery, (4) define capabilities as business-aligned workflows, not just endpoints, and (5) provide safe, governed execution environments. It reframes readiness as a lifecycle discipline and shows how to turn pilots into durable, governed capabilities using open standards that protect sovereignty. Takeaways: The five-step API enablement stack; why API exposure alone doesn't scale; how standards-based workflows and governance make AI repeatable in production; patterns for measuring and improving API AI-readiness at scale.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4495,
  "day": "2026-12-03",
  "type": "Keynote",
  "start": "09:45",
  "end": "10:10",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Prototyping tomorrow's API: inside IGN's proof-of-concept factory",
  "speakers": [
   {
    "name": "Valentine Bruyerre",
    "title": "Head of innovation",
    "company": "IGN - French Mapping Agency"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability"
  ],
  "agent_readiness_dimensions": [
   "mcp_server"
  ],
  "why": "A public operator prototyping AI access to its authoritative APIs.",
  "abstract": "Public institutions hold authoritative data, yet AI assistants cannot reliably answer questions without access to it. How can a public operator make its APIs usable by AI, and test such ideas quickly before committing to large-scale development? This talk presents IGNfab, the proof-of-concept and demonstrator factory of the French National Institute of Geographic and Forest Information (IGN). A team of engineers works with ministries, public operators and local authorities, with a light, iterative process: define the need, scope the experiment, build a prototype, then deliver a solid basis for deciding whether to scale up. We will use one of IGNfab's experiments as a case study. Two open source MCP (Model Context Protocol) servers, Geocontext and Carto, connect the Géoplateforme, the infrastructure behind cartes.gouv.fr, to any AI assistant. Users can query reference geodata in natural language, for example to locate every school in a municipality or find homes within 3 km of a point, and generate maps without writing a GIS query. Why it matters: public data is only valuable if people and machines can use it. Other experiments are under way, such as 3D territory visualization and AI-based flood mapping, showing how rapid prototyping helps API providers prepare for the AI era.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4590,
  "day": "2026-12-03",
  "type": "Keynote",
  "start": "10:05",
  "end": "10:40",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "How green is AI?",
  "speakers": [
   {
    "name": "Nina Benoit",
    "title": "Director of Sustainability",
    "company": "Brightest"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI's carbon, water and mineral footprint.",
  "abstract": "In this talk, I will explore both the negative and positive sides of AI, looking into the carbon, water, and mineral footprint, but also different solutions that AI can provide in terms of biodiversity conservation, waste sorting and agriculture. This talk will try to answer the question: How green is AI? in a balanced way giving tips on how to reduce your footprint.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4738,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Houston, We Have a Partial Response: GraphQL Error Handling",
  "speakers": [
   {
    "name": "Željko Kozina",
    "title": "Technical Architect",
    "company": "Chronomill"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "error_semantics",
   "mcp_tool_quality"
  ],
  "why": "GraphQL partial-response errors and workflow-backed MCP tools.",
  "abstract": "The difference is measurable. This talk walks through how workflow-backed MCP tools affect task completion rates, token efficiency and predictable execution cost compared to operation-by-operation exposure, and in a manner that can be verified across models, so the improvements hold regardless of which LLM is doing the work. Evals become more possible which turns \"it worked in testing\" into \"we know it works at the boundary, consistently and within budget.\"",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4496,
  "day": "2026-12-03",
  "type": "Keynote",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "How API and AI Tools work together across Netherlands Government",
  "speakers": [
   {
    "name": "Dimitri van Hees",
    "title": "",
    "company": "Geonovum"
   }
  ],
  "facets": [
   "discoverability",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; API and AI tooling across Dutch government.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4352,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "How did the agents spread within Carrefour?",
  "speakers": [
   {
    "name": "Guillaume Blaquiere",
    "title": "Group Data Architect",
    "company": "Carrefour"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Internal agent adoption story at Carrefour; no API surface.",
  "abstract": "The agents have started to conquer the world in 2025. Carrefour has been invaded too! And it changes the way of working, interacting and our velocity on different topics. Come to discover how it started, the different iteration and team impacted, and the plans for the future. Learn the different steps, our pain points and the quick wins. And get back to work to start the agentic conquest in your team!",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4363,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "How much does your AI cost, really ?",
  "speakers": [
   {
    "name": "Alex Hayem",
    "title": "CEO / Founder",
    "company": "MIRMI AI"
   },
   {
    "name": "Enzo Sad-Eddine",
    "title": "Group Principal IA",
    "company": "Thales"
   }
  ],
  "facets": [
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "Hidden AI production costs; only the pricing piece of Access Clarity touches it.",
  "abstract": "What does AI really cost in production? Hidden costs, scaling and architecture choices: 25 minutes to understand what can multiply — or reduce — the bill.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4633,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Tokens are not compute: Getting more from LLMs by using them less",
  "speakers": [
   {
    "name": "Jens Neuse",
    "title": "CEO",
    "company": "WunderGraph"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality",
   "operation_distinctiveness"
  ],
  "why": "Every endpoint as a tool bloats context; fewer, better tools.",
  "abstract": "Most agents today use APIs the same way: every endpoint becomes a tool, every tool definition sits in the context window, and every call goes through the model, one at a time, with each result read back in. We've turned a language model into a CPU and tokens into our unit of compute. It's slow, expensive, and the agent picks the wrong operation more often than anyone admits. The problem isn't MCP, it's how we're looping over tool calls. Over the last few months, we've tested a different pattern against APIs like GitHub, Linear, Pylon and Circleback: a frontier model writes code against typed SDKs, and we run that code in a sandbox. The model never sees API credentials, and the context window stays clear of intermediate results. If a step needs to classify or summarize, the script calls a smaller model. It's just another API call. The most interesting part: we can now iterate on the script until the outcome is what we want, like any other program. Running it again costs cents, because the frontier model is out of the loop, and it finishes much faster. In this talk, we'll walk you through our experiments, including the ones that failed, and what they taught us about how agents actually call APIs. We're also open-sourcing our solution, so you'll walk away with a concrete example to try, a way to measure cost and efficiency for agentic tasks, design choices that make your REST, GraphQL and MCP APIs not just callable but agent-friendly, and a sense of where the model belongs in the loop and where it doesn't.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4671,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "No tool has ever fixed a testing problem",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; testing culture, not API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4371,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Risk is Governed. Data Resilience is Architected.",
  "speakers": [
   {
    "name": "Siham Boutayeb",
    "title": "Sr Principal Solutions Engineer",
    "company": "Odaseva"
   }
  ],
  "facets": [
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [],
  "why": "Board risk governance vs data-system resilience architecture.",
  "abstract": "A reflection on the alignment — or misalignment — between board-level risk governance and the architectural reality of data systems. Risk is discussed in boardrooms. Resilience is implemented in systems. But too often, the two operate in parallel universes. What does it mean to treat resilience not as a technical feature, but as a strategic discipline? How do we translate regulatory accountability into architectural patterns? Where does operational resilience truly live — in policy documents, or in recovery-tested systems? And how do we bridge the gap between governance language and data-layer implementation? This session explores how to structure resilience as a strategic capability — one that is designed, measured, tested, and owned.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4696,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "10:15",
  "end": "10:40",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Taking the risk out of AI-written software, without slowing down",
  "speakers": [
   {
    "name": "Vincent Verdet",
    "title": "Senior Enterprise Architect",
    "company": "Independent"
   }
  ],
  "facets": [
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "Guarding AI-written code in enterprises; internal practice.",
  "abstract": "Most industries outside Big Tech are wary of AI-written code, and with reason: more developers distrust its output than trust it. Yet the speed is real. The question for IT leaders is how to take the speed without the risk. I answered it by building Turbo EA, an open-source Enterprise Architecture platform, as one architect working with AI coding agents: about 400,000 lines of code of enterprise-grade software in eight months, covered by more than 12000 automated tests. AI is probabilistic, a turbulent flow. Software has to be laminar. This talk shows the five rules that straighten the flow without slowing it down: a stack the AI knows, written guidelines, reuse principle, security layers and automation. It also shows the CI/CD pipeline that enforces them, itself generated by AI. You will leave with a risk framework you can explain to your organization.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4661,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "AI Sovereignty Is an Architecture Problem, Not a Model Choice - Keeping Data, Policy and Execution Under Your Control",
  "speakers": [
   {
    "name": "Giona Granchelli",
    "title": "Chapter lead Java & VueJs",
    "company": "ABN Amro"
   }
  ],
  "facets": [
   "regulatory",
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Sovereignty over data, policy and execution in agentic systems.",
  "abstract": "AI sovereignty is often reduced to one question: where does the model run? But running an open model locally does not automatically give an organisation control over its AI system. As AI applications become more agentic, models are no longer just generating text. They access data, call APIs, choose tools and trigger actions. This creates a much broader sovereignty problem: who controls what data can leave the system, which models may process it, what actions are allowed, and which policies are enforced before execution? This talk explores AI sovereignty as an architectural problem rather than a model choice. Drawing from lessons learned while building TramAI, an open-source Kotlin runtime for governed AI workflows, I will show how to separate policy, data boundaries, model routing and tool permissions from the models themselves. We will look at architectures where sensitive workloads can remain local, cloud models can still be used when policy allows it, dangerous actions can require explicit approval, and every important decision remains observable and auditable. The goal is not to eliminate cloud AI. It is to build systems where models, providers and infrastructure can change without giving up control over data, policy or execution.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4647,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Your Domain Model Is Your API Spec — and Your First Prompt",
  "speakers": [
   {
    "name": "Dr. Annegret Junker",
    "title": "Chief Software Architect",
    "company": "codecentric AG"
   }
  ],
  "facets": [
   "contract_quality",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Domain model as the API spec that grounds LLM generation.",
  "abstract": "Most teams have now tried handing an LLM a vague description and asking for an API. The results are familiar: plausible-looking endpoints, invented fields, events that don't match anything real, and bugs that take hours to trace. The easy reaction is to blame the model. That's the wrong angle. LLMs don't produce wrong APIs because they're not capable — they produce wrong APIs because they have no access to your domain. The fix is something API teams can already recognize: a precise specification. This talk shows that the collaborative modeling practices used to build shared understanding — Domain Storytelling and EventStorming — double as a remarkably good way to derive that specification, across both styles of API. EventStorming's command/event/policy structure maps almost one-to-one onto an API surface: commands become synchronous REST operations, domain events become asynchronous channels and messages, and policies become the orchestration and business rules that connect them. Bounded contexts become clean service and product boundaries. The result is a paired OpenAPI and AsyncAPI specification that is, in effect, a machine-readable domain model. Using CookWithUs, a recipe-sharing platform modeled and built across three iterations, we walk a concrete pipeline: model the domain, derive the synchronous REST surface and the event-driven async surface together, and split both specs by bounded context. We show exactly what a well-formed spec gives an LLM — and exactly where generation breaks down when the spec is incomplete, ambiguous, or missing its events. We show how a sharper Ubiquitous Language produces a measurably better OpenAPI document, a cleaner set of AsyncAPI channels, and a frontend the model can generate without inventing a single field. The central claim is simple: every naming decision, every event you make explicit, every scoping choice travels all the way into the generated code. A spec is only as good as the language behind it — and async APIs in particular punish vague language, because an event nobody named precisely is an event nobody can consume reliably. This is not a talk about replacing API designers with AI. It's about what happens when rigorous API specification meets a powerful new tool, and why the modeling work always comes first.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4582,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Exclusive presentation: Worldwide AI environmental footprint study (English version)",
  "speakers": [
   {
    "name": "Auban Derreumaux",
    "title": "Director, Treasurer, LCA Analyst in Digital",
    "company": "Association Green IT"
   },
   {
    "name": "James Martin",
    "title": "Founder",
    "company": "BetterTech"
   },
   {
    "name": "Laure Alfonsi",
    "title": "Consultante Numérique Durable",
    "company": "Zeb et Web"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Worldwide AI environmental footprint study.",
  "abstract": "Association Green IT published, and showcased at GreenIO last year, the first version of its study about AI environmental footprint at world scale, based on multi-criteria LCA/ISO 14040 norm, and European PEF standard with up to 16 indicators, going beyond GreenHouse Gas emissions (fossil resources, water use, human being and ecosystem sanitary impacts). This year, we deeply updated the study improving data quality and infra model and the forecast to 2030. And, exclusive to GreenIO, we will announce the first publication of the complete study report in English, for wider audience.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4668,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Generative AI Plateforme Programm, transform the public sector",
  "speakers": [
   {
    "name": "Adrien Benamira Carrie",
    "title": "Generative AI Plateforme Manager for the DGFiP",
    "company": "Direction Generale des Finances Publiques"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; public-sector GenAI programme, no API signal in the title.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4881,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Let’s Help Agents to Navigate the API and the Data World",
  "speakers": [
   {
    "name": "Dr. Simon Harrer",
    "title": "Co-founder and CEO",
    "company": "Entropy Data"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; helping agents navigate APIs and data.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4548,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "In the search of Datacenters and their footprint with the DCWatch project",
  "speakers": [
   {
    "name": "Benoit Petit",
    "title": "Co-founder",
    "company": "Hubblo"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Datacenter footprint measurement.",
  "abstract": "How can we assess their current and future footprint of Datacenters, with the advent of AI? To answer this question, we will address four points: What generative AI does to data centers The global vs the local impacts Results and methodology used for the ADEME 2060 prospective study in France and for the DCWatch project data collection Demonstration of DCWatch: an open and collaborative research project for Datacenters and their footprint",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4645,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Fast AI needs rules: governing agents, models, and tokens",
  "speakers": [
   {
    "name": "Raphaël  Velfre",
    "title": "Strategic Account Executive",
    "company": "Kong Inc"
   }
  ],
  "facets": [
   "operational_transparency",
   "agent_readiness",
   "accountability"
  ],
  "agent_readiness_dimensions": [
   "rate_limit_signal",
   "agent_identity_declared"
  ],
  "why": "Governing who calls which model with what data at the traffic layer.",
  "abstract": "Everyone’s shipping AI. Few can say who calls which model, with what data. Raphael Velfre makes the case for governing AI at the traffic layer, so speed and control stop being a tradeoff.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4599,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Claude to Copilot - Measuring AI SaaS Emissions",
  "speakers": [
   {
    "name": "William Tinney",
    "title": "CTO & Founder",
    "company": "Greenpixie"
   },
   {
    "name": "Joshua Horswill",
    "title": "Data Scientist",
    "company": "Greenpixie"
   }
  ],
  "facets": [
   "outside",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI SaaS emissions from provider usage reports; a usage-disclosure gap.",
  "abstract": "The first thing you need, to get activity based emissions data for AI is… well, your activity! AI providers monitor usage in many different ways and GPX AI supports all of them. In this workshop we'll cover the specific usage reports available to you from M365 Copilot and the full end-to-end journey of how GPX AI can turn these reports into activity-based emissions data.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4364,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Your Domain Model Is Your API Spec - and Your First Prompt",
  "speakers": [
   {
    "name": "Dr. Annegret Junker",
    "title": "Chief Software Architect",
    "company": "codecentric AG"
   }
  ],
  "facets": [
   "contract_quality",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Domain model as the API spec that grounds LLM generation.",
  "abstract": "Most teams have now tried handing an LLM a vague description and asking for an API. The results are familiar: plausible-looking endpoints, invented fields, events that don't match anything real, and bugs that take hours to trace. The easy reaction is to blame the model. That's the wrong angle. LLMs don't produce wrong APIs because they're not capable — they produce wrong APIs because they have no access to your domain. The fix is something API teams can already recognize: a precise specification. This talk shows that the collaborative modeling practices used to build shared understanding — Domain Storytelling and EventStorming — double as a remarkably good way to derive that specification, across both styles of API. EventStorming's command/event/policy structure maps almost one-to-one onto an API surface: commands become synchronous REST operations, domain events become asynchronous channels and messages, and policies become the orchestration and business rules that connect them. Bounded contexts become clean service and product boundaries. The result is a paired OpenAPI and AsyncAPI specification that is, in effect, a machine-readable domain model. Using CookWithUs, a recipe-sharing platform modeled and built across three iterations, we walk a concrete pipeline: model the domain, derive the synchronous REST surface and the event-driven async surface together, and split both specs by bounded context. We show exactly what a well-formed spec gives an LLM — and exactly where generation breaks down when the spec is incomplete, ambiguous, or missing its events. We show how a sharper Ubiquitous Language produces a measurably better OpenAPI document, a cleaner set of AsyncAPI channels, and a frontend the model can generate without inventing a single field. The central claim is simple: every naming decision, every event you make explicit, every scoping choice travels all the way into the generated code. A spec is only as good as the language behind it — and async APIs in particular punish vague language, because an event nobody named precisely is an event nobody can consume reliably. This is not a talk about replacing API designers with AI. It's about what happens when rigorous API specification meets a powerful new tool, and why the modeling work always comes first.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4338,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "A Valid API Call, An Invalid Outcome - Accountability in the Age of AI Agents",
  "speakers": [
   {
    "name": "Afsha",
    "title": "DevOps Engineer",
    "company": "Capgemini"
   }
  ],
  "facets": [
   "accountability",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "reversibility_documented",
   "dry_run_mode"
  ],
  "why": "Valid API calls by agents with invalid outcomes; who answers for them.",
  "abstract": "AI agents no longer only recommend actions. They act through APIs (application programming interfaces). They can update records, provision infrastructure and trigger business workflows, potentially without a person reviewing each decision. APIs are becoming the hands through which AI agents affect people and organisations. Technical reliability matters, but it is no longer enough. A valid API call can create an invalid human outcome. Authentication, authorisation and schema validation can show that a request is permitted and correctly formed. They cannot tell us whether the action is wise, fair or ethical. This talk examines two responsibilities we cannot outsource to AI: ethical judgement and accountability, and human connection and transparency. What happens when a technically valid action produces an unacceptable outcome? When something goes wrong, can the person affected understand what happened, challenge the decision and reach a human who can put it right? I will introduce my SCOPE framework, a practical checklist for deciding what an agent should be allowed to do before it ever gets access. By the end, attendees will be well placed to shape agentic API systems that are secure and reliable while remaining transparent, accountable and deserving of the autonomy we give them. This is a beginner-friendly talk designed for both technical and non-technical audiences. No coding or API experience is needed. Whether you build these systems, lead the teams that do, or are simply affected by them, everyone is welcome.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4672,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Self-hosted document extraction, served over API and MCP",
  "speakers": [],
  "facets": [
   "agent_readiness",
   "open_source"
  ],
  "agent_readiness_dimensions": [
   "mcp_server"
  ],
  "why": "No abstract; self-hosted extraction served over API and MCP.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4374,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:00",
  "end": "11:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "The API Is Not the Authority: Governing Autonomous AI",
  "speakers": [
   {
    "name": "Pedro Barbas",
    "title": "Founder, Chief Architect",
    "company": "TrustGate Sovereign"
   }
  ],
  "facets": [
   "accountability",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "delegated_identity",
   "consent_identity",
   "agentic_commerce"
  ],
  "why": "Authority for consequential agent actions beyond auth and permissions.",
  "abstract": "Agentic AI is moving from generating recommendations to taking consequential actions: calling APIs, initiating payments, changing infrastructure, placing orders, and triggering enterprise workflows. That creates a control problem that authentication, API permissions and conventional policy engines do not fully solve. Knowing who an agent is, which API it can call, or what scope it has been granted is not the same as proving that this exact action is authorised, through this route, against this target, under the enterprise’s current state. This talk explores an execution-control model for autonomous systems built around a simple distinction: access is not authority. Before execution, the system must bind current authority to the precise protected action and constrain how that action can occur. After execution, it must establish what actually happened in the external world — and fail safely when the outcome is incomplete, ambiguous or contradictory. Using examples from payments, procurement and infrastructure automation, we will examine where existing API and identity controls stop, what additional runtime evidence becomes necessary, and how enterprises can prevent autonomous execution from silently amplifying authority during retries, failures and recovery. The goal is a practical architecture for letting AI agents act without giving them more authority than the enterprise intended.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4586,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:25",
  "end": "11:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Right to Repair vs. Intellectual Property: Hardware’s Second Life",
  "speakers": [
   {
    "name": "Carlyne Sevestre",
    "title": "Attorney-at-law",
    "company": "Leca Sevestre Avocats"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Right to repair vs IP for hardware.",
  "abstract": "Europe's push to reshore hardware and extend device lifespans runs straight into Intellectual Property law. The Right to Repair Directive (2024/1799) requires manufacturers to offer repair and spare parts, while the EU design reform introduces a \"repair clause\" opening the market for visible spare parts. Yet refurbishers and resellers of branded hardware remain exposed: limits of trademark exhaustion, use of the original brand on repaired goods, counterfeiting claims. This talk maps where IP still blocks a genuine second life and how to build repair-friendly IP strategies.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4556,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "11:25",
  "end": "11:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Leave no stone unturned: assessing the full impacts of your organisation’s digital tech using DIST",
  "speakers": [
   {
    "name": "Oliver Cronk",
    "title": "Founder",
    "company": "Architect Tomorrow | Green Web Foundation"
   },
   {
    "name": "Hannah Smith",
    "title": "Director of Operations",
    "company": "Green Web Foundation"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Full digital-tech impact assessment beyond carbon.",
  "abstract": "Frugal, sustainable and efficient tech must account for the entire IT supply chain, from manufacturing and operational use to e-waste and end-user outcomes. This means looking beyond carbon to include other environmental impacts, such as water use and biodiversity, to build truly resilient, frugal and fair tech. In this interactive workshop, we’ll explore how DIST, an open-source project stewarded by Green Web Foundation, provides accessible, standards-aligned guidance for more comprehensive digital tech reporting. We’ll also explore ways to contribute to the project, whether you’re new to digital sustainability or an experienced practitioner.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4552,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:25",
  "end": "11:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Building Green IT from the inside: ABB's Journey from awareness to reality",
  "speakers": [
   {
    "name": "Thomas Mosser",
    "title": "IT Sustainability Lead",
    "company": "ABB"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "ABB's internal Sustainable IT journey.",
  "abstract": "How does a global industrial group of 100,000+ employees turn good intentions into sustainable IT operations? This talk shares ABB's Sustainable IT journey from the inside. Climate Fresk workshops and an IT footprint assessment ignited the change and set foundations for a data-driven roadmap aligned with ABB's sustainability goals. Next came shared governance and flagship initiatives like PC lifespan extension, Cloud GreenOps, local initiatives. We'll share what worked, what didn't, and what comes next, from eco-design to IT for Sustainability. Practical lessons for newcomers and experts alike.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4673,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "AI writes the code, so who writes the proof? E2E testing in AI-assisted development",
  "speakers": [
   {
    "name": "Pier-Jean Malindrino",
    "title": "CTO",
    "company": "Scub"
   }
  ],
  "facets": [
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; E2E testing of AI-written code.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4665,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "From expert judgments to LLM quality signals",
  "speakers": [
   {
    "name": "Arianna Bonazza",
    "title": "Data scientist",
    "company": "Publicis Groupe"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Evaluating LLM quality with expert judgments; no API surface.",
  "abstract": "Large Language Models are increasingly used to support sensitive decisions in advertising, such as detecting greenwashing risks or assessing inclusive representation. But evaluating LLMs in these domains with classic data science is difficult: the “right answer” is rarely a single fact and often depends on expert interpretation. This creates a measurement gap—teams need quantitative signals to choose and improve models, while still respecting the qualitative nature of expert judgment, especially when labeled data is scarce. We built the Positive AI LLM Arena, a Streamlit-based evaluation platform that turns expert review into a guided, scalable, and rigorous benchmark. By structuring the workflow and reducing reviewer effort, the Arena transforms data collection from a tedious task into a smoother (and even enjoyable) process, helping teams collect high-quality judgments at scale. The key innovation is gamified, blind comparison. In the Analysis Arena, experts review an asset, replicate the reasoning expected from the LLM, label principles using operational categories, and then compare their decisions to anonymized, side-by-side outputs from different models. A lightweight progress scoring layer sustains consistent, high-standard reviews and produces continuous evaluation signals and decision-grade ground truth. This talk explains how ML user experience patterns (gamification) and human-in-the-loop design can reduce bias, improve consistency, and generate reliable benchmarks that directly inform model selection, performance tracking, and reliability improvements for responsible communication use cases.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4355,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "The Hidden Complexity of Aggregated APIs: Lessons from Optimizing a Global Hotel Distribution Platform",
  "speakers": [
   {
    "name": "Nermine Reyadh",
    "title": "Director of optimisation",
    "company": "Nuitée"
   }
  ],
  "facets": [
   "operational_transparency",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "pagination_ratio",
   "error_semantics"
  ],
  "why": "Aggregating third-party supplier APIs whose real behaviour diverges from contracts.",
  "abstract": "APIs look clean on paper: structured contracts, predictable responses, elegant documentation. But when you aggregate dozens of third-party suppliers into a single global platform, the reality is far messier. In this talk, I’ll share practical lessons from optimizing a large-scale hotel distribution API powering search, prebook, and booking flows across multiple suppliers and markets. Behind every “200 OK” lies a complex ecosystem of inconsistent rate structures, mapping mismatches, duplicate properties, timeout constraints, rate violations, and fragile upstream systems. We’ll explore what really happens when you try to guarantee performance and reliability in an environment you don’t fully control. How do you handle supplier instability without degrading user experience? How do you design around strict booking SLAs (search → prebook → book) under heavy traffic? How do you balance latency, accuracy, cost, and margin? This session is not about theory or API design patterns in isolation. It’s about operating aggregated APIs in the real world — where commercial impact, infrastructure constraints, and ecosystem chaos intersect. If you work with third-party integrations at scale, this talk will give you practical frameworks and battle-tested insights you can apply immediately.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4498,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Register Once, Integrate Everywhere: FAPI 2.0 and a Federated Authorization Foundation for Public-Sector APIs",
  "speakers": [
   {
    "name": "Alexander Hoose",
    "title": "Enterprise Architect im Föderalen IT-Architekturmanagement",
    "company": "Föderale IT-Kooperation (FITKO)"
   }
  ],
  "facets": [
   "access_clarity",
   "regulatory",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "dynamic_client_registration",
   "auth_clarity",
   "protected_resource_metadata"
  ],
  "why": "FAPI 2.0 and federated registration for public-sector APIs.",
  "abstract": "Public-sector APIs are becoming critical infrastructure for digital governments. Yet in federated environments, services define their own authorization models, registration processes and security requirements. This leads to duplicated integration effort, inconsistent permission models and fragmented security across the public-sector API landscape. The Federal API Authorization Infrastructure project, developed by Saxony-Anhalt and FITKO on behalf of the German IT Planning Council, addresses this challenge by defining uniform security requirements and a target architecture for federal API authorization. The approach is based on established open standards such as OAuth 2.0, OpenID Connect, the FAPI 2.0 Security Profile and AuthZEN, rather than introducing a bespoke authorization scheme. The talk presents the architectural vision, scope and key design decisions. We explain how the architecture structures API provision, integration and usage across federal base services, and how the Federal Platform Directory acts as a central self-service component for registration, API-client onboarding and managing authorization attributes and policies. We will discuss selected trade-offs, including sender-constrained tokens, client authentication, signed software statements, attribute-based authorization and the balance between centrally governed policies and decentralized execution and enforcement. The session also shows how reusable architecture artifacts, security requirements and migration paths support operators, developers and security teams in building more secure, interoperable and economically viable API ecosystems in federated public administration. All artifacts — security requirements, requirements documents, architecture decision records, and the target architecture — are published openly on OpenCode: https://gitlab.opencode.de/sachsen-anhalt/mid/foederale-apiautorisierungsinfrastruktur/-/ blob/main/README.md",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4739,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Foundational JSON Schema",
  "speakers": [
   {
    "name": "Ben Hutton",
    "title": "Staff Backend Engineer",
    "company": "Spoke / JSON Schema TSC"
   }
  ],
  "facets": [
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "JSON Schema fundamentals behind well-formed contracts.",
  "abstract": "New to JSON Schema? Think you know enough JSON Schema? Think AI can do all your JSON Schema for you? When things go wrong, you'll need a good foundational understanding to make sure it does what you think it does, or debug why it's not working like it should.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4349,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Data Science : the multibillons dollars failures",
  "speakers": [
   {
    "name": "Marie Sacksick",
    "title": "Head of Market Intelligence",
    "company": "Probabl"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Data science project failures; no API relevance.",
  "abstract": "At a time where more and more budget is poured into AI, and the infamous rule of 80% data science projects which don't make it to production (or are cut early) is still relevant, let's investigate the why! We’ve listed 126 public cases in which a data science project led to an economical disaster. In this talk, we want to have a look about the why these projects failed, how much they cost to the company, if the failure could have been avoided, and if so, how. As failures that are public are necessarily huge scale, and with end-user impact, they do not cover the whole reality. It will be enriched with failures known by the speaker, either because they are personal, or acquaintances stories.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4506,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:30",
  "end": "11:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "From click to terminal: API engineering in the age of agents",
  "speakers": [
   {
    "name": "Anthony Viard",
    "title": "Senior Developer Advocate",
    "company": "Postman"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Agents running API engineering: contract, docs, tests, monitoring, CI/CD.",
  "abstract": "We can define API engineering as the set of practices that keep an API alive independently of the code that implements it: its contract, documentation, tests, monitoring, CI/CD, and so on. Doing API engineering with agents means making them autonomous across that entire chain: reading a contract, analyzing risk, running tests, interpreting results. But without proper framing and tooling, an agent can easily drift or hallucinate, pulling us away from our goal. That risk is all the more real today, as an agent is already capable of producing and reviewing code on its own, in natural language. This is why we need to design our tools not only for the developers who steer the agent, but for the agent itself. In this talk, we'll look at how to improve \"\"Agent Experience\"\" by designing tools that an agent can understand and use on its own, then explore the work the teams did to bring the Postman CLI, the plugin, and the skills to the terminal. We'll then see how my Claude Code agent uses them to build an API from scratch and integrate it into CI/CD: running an analysis, setting up a quality gate that ensures the API stays understandable to an agent, or verifying that a change doesn't break anything elsewhere in the organization.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4584,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Is Digital Sovereignty Sustainable?",
  "speakers": [
   {
    "name": "Georgina Macrae",
    "title": "CAT Partnerships Organiser",
    "company": "ClimateAction.Tech"
   }
  ],
  "facets": [
   "outside",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "Sovereignty versus sustainability trade-offs.",
  "abstract": "Is local always greener? As organisations pursue digital sovereignty, bringing data, infrastructure and AI closer to home may increase control, but it doesn't automatically improve sustainability. This talk explores where sovereignty and ESG reinforce each other, and where they collide: from energy and water to data privacy, AI governance and local communities. We'll ask how architects can make technology choices that are not only sovereign, but genuinely sustainable and responsible.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4595,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "11:50",
  "end": "12:15",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "La Poste: Establishing Consistent Footprint Measurement Across Multi-Entity Environments",
  "speakers": [
   {
    "name": "Matthieu Poulard",
    "title": "CEO",
    "company": "Aguaro"
   },
   {
    "name": "Annick Lebreton",
    "title": "Référente Sobriété Numérique",
    "company": "La Poste"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Footprint measurement across La Poste entities.",
  "abstract": "As the first French public-sector \"entreprise à mission\" (purpose-driven company), commitment is embedded in the DNA of La Poste, from the Group level down to all its subsidiaries. As part of its Level 2 Sustainable IT policy, the Group is leading the implementation of IT footprint measurement across a complex landscape of multiple entities and diverse tools. In an organization of over 200,000 employees, ensuring long-term accuracy, traceability, and auditability requires a robust and resilient approach. Join us to discover how La Poste leverages Aguaro to consolidate data flows and guarantee measurement consistency, while empowering its Sustainable IT teams to remain autonomous.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4551,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "11:50",
  "end": "13:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Sustainable Digital Challenge : les pitchs des équipes",
  "speakers": [
   {
    "name": "Nicolas Drouet",
    "title": "CEO",
    "company": "The Green Compagnon"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Team pitches for a sustainability challenge; no API relevance.",
  "abstract": "Les équipes du Sustainable Digital Challenge présentent leur projet devant le jury : 5 minutes de pitch par équipe, suivies des questions du jury. Équipes : SNCF, Servier, Enedis, Schneider Electric. Organisé par The Green Compagnon.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4740,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "GraphQL with AWS AppSync - Building a Product API at HEMA",
  "speakers": [
   {
    "name": "Ivan Jančić",
    "title": "Solutions Architect",
    "company": "Brighting"
   },
   {
    "name": "Bart Rijksen",
    "title": "",
    "company": "HEMA"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "GraphQL product API adoption on AppSync.",
  "abstract": "I will talk about \"pioneer phase\" of GraphQL adoption at HEMA and using AWS AppSync for the composable commerce migration, where I will share the raw and honest lessons learned when introducing a new tech stack to a 100-year-old Dutch icon.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4664,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Continuous Expert-in-the-loop Agentic AI Validation",
  "speakers": [
   {
    "name": "Quincy Hsieh",
    "title": "Senior Solution Architect",
    "company": "Milliman"
   },
   {
    "name": "Bertille Tierny",
    "title": "Actuary Consultant",
    "company": "Milliman"
   },
   {
    "name": "Muhammad Umer",
    "title": "AI engineer",
    "company": "Milliman"
   }
  ],
  "facets": [
   "agent_readiness",
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "Expert-in-the-loop validation of agentic AI output.",
  "abstract": "Unlike traditional software development, where output correctness can often be verified against predefined expected integration tests, validating an agentic AI requires sustained expert involvement to define what “good” means in context, to review ambiguous outputs, and to translate domain judgment into measurable evaluation criteria. Achieving this in practice, however, requires more than technical evaluation: enterprise deployment depends on aligning domain experts and AI engineers so that model choices reflect real business requirements. In many organizations, the gap between these two groups is where AI projects quietly fail when business accuracy diminishing by time. In this session, we introduce how Milliman streamline collaboration between domain experts and AI engineers, to accelerate validation cycles and to continuously improve the business accuracy from agents.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4653,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Google Talk",
  "speakers": [
   {
    "name": "Christophe Lalevée",
    "title": "Apigee Customer Engineering Lead",
    "company": "Google Apigee"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; sponsor talk with no stated topic.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4323,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "The Automation Paradox",
  "speakers": [
   {
    "name": "Kate Astrid",
    "title": "Senior Product Engineer",
    "company": "Cogram"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Human skills under automation; no API relevance.",
  "abstract": "When AI does the easy work, humans get the hard parts. Automation removes repetitive tasks, but it also changes the skills humans need. As AI takes over coding, analysis, writing, and decision-making, we're increasingly left with ambiguity, exceptions, judgment, and responsibility. This talk explores what automation does to human cognition and expertise. What happens when we stop practicing skills because machines do them for us? Why do we trust automated answers? And as AI becomes better at producing answers, does human expertise become more or less valuable?",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4402,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Fostering AI skills and capabilities in large enterprises",
  "speakers": [
   {
    "name": "Saheed Abiola Lasisi",
    "title": "Executive Director - Principal API Architect",
    "company": "JP Morgan Chase"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; enterprise AI skills building, no API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4945,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Context management in large organisations : How to Make better AI decisions",
  "speakers": [
   {
    "name": "Alex Hayem",
    "title": "CEO / Founder",
    "company": "MIRMI AI"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Enterprise context management for AI tools; no public API surface the rubric reads.",
  "abstract": "Most companies are not failing at AI because they lack powerful models. They are failing because even today’s leading enterprise AI tools — from Enterprise Copilot to GPT Enterprise — remain weak at reasoning over context. When a user asks, “Have we done this before?”, most AI systems tend to answer in a binary way: yes or no. But in large organisations, the real question is rarely whether something happened before. It is whether the previous case was truly comparable: same type of counterparty, same risk level, same commercial upside, same legal constraints, same exception logic. This is where context management becomes critical. Enterprise AI cannot simply retrieve documents through RAG and assume that similarity equals relevance. To understand context properly, organisations need more efficient models, but also a stronger interconnection of internal data points — across ERPs, CRMs, CLMs, document management systems, policies and knowledge bases. APIs play a central role here: they allow AI systems to access the right signals, from the right systems, at the right moment, to understand when a precedent should apply — and when it should not. The issue is amplified by current AI user maturity. Employees often use short, direct prompts, which limits contextualisation. But overly complex prompts are not a reliable solution either: they increase the risk of hallucination and misleading outputs. Drawing from my experience building API aggregation infrastructure at The Good Seat and internal contract memory at Mirmi AI, this talk will explore how large organisations can design better context systems so AI can support safer, more consistent decisions.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4674,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:00",
  "end": "12:25",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Testing enterprise PLM APIs with Karate: 3DExperience, Aras and custom web services",
  "speakers": [
   {
    "name": "Hadrien Guiducci ",
    "title": "Architecte Cloud/DevOps",
    "company": "CIMPA PLM Services"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; API testing of PLM web services with Karate.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4559,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "12:15",
  "end": "12:40",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "The Responsible AI Canvas 2.0 - in depth evaluation of AI use-cases",
  "speakers": [
   {
    "name": "Thorsten Jonas",
    "title": "Founder / Digital Sustainability Trailblazer",
    "company": "SUX - The Sustainable UX Network"
   },
   {
    "name": "Anna Forlati",
    "title": "Head of Digital Sustainability",
    "company": "TeamSystem"
   }
  ],
  "facets": [
   "accountability",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Responsible AI canvas for evaluating use cases; governance of AI, not of APIs.",
  "abstract": "After the introduction of the Responsible AI Canvas last year, Anna Forlati and Thorsten Jonas now present the updated version that adds a new step to the AI use-case evaluation. Now it is not only possible to identify risks and opportunities for organisations as well as negative and positive impacts for the surrounding ecosystem but afterwards make in depth evaluations of each identified impact by adding a new layer inspired by the double materiality principles. (Note: this is a workshop, it would also be possible to it as a talk)",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4583,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:15",
  "end": "12:40",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Can Africa Build Green AI Infrastructure? Lessons from the Democratic Republic of Congo",
  "speakers": [
   {
    "name": "Narcisse Mbunzama",
    "title": "Senior AI Researcher",
    "company": "Digital Security Group"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Green AI infrastructure in the DRC; sustainability, nothing the rubric measures.",
  "abstract": "Africa’s AI ambitions depend on digital infrastructure that is both scalable and sustainable. Using the Democratic Republic of Congo as a case study, this talk explores the infrastructure challenges behind Green AI—from energy access and data centers to connectivity, compute capacity and cooling. It examines how efficient AI models, distributed infrastructure, renewable energy and locally adapted architectures could enable Africa to expand AI while reducing its environmental footprint.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4341,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Less Status, More Signal: How AI Agents Changed Delivery Management",
  "speakers": [
   {
    "name": "Alexis Revue",
    "title": "Senior Staff Delivery Manager",
    "company": "Ledger"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Internal agent platform for delivery management; no public API signal.",
  "abstract": "Most delivery teams do not lack information. They lack usable signal. We built an AI agent platform specialized in delivery management, adapted to our context and connected to our everyday tools. The ambition was simple: reduce the coordination burden on Delivery Managers by delegating repetitive work such as status collection, meeting preparation, action tracking, risk detection, dependency visibility, and project context retrieval. But the real lesson was sharper: AI agents do not just improve delivery management; they expose it. They reveal where reporting creates noise, where rituals lack decisions, where ownership is unclear, and where teams confuse visibility with control. This talk is a candid field report on what changed when we asked agents to take over a significant part of delivery operations. I will share where they created leverage, where they failed, and what we had to rethink to turn delivery noise into signal without outsourcing judgment, trust, or accountability.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4675,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "I made the call: what I chose, and what I got wrong",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; a personal retrospective with no stated API topic.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4348,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "When the Buyer is an Agent: How AI Intermediaries Disrupt eCommerce Acquisition",
  "speakers": [
   {
    "name": "Katia Maevskaia",
    "title": "Commercial & Revenue Growth Leader",
    "company": "Snap Inc"
   }
  ],
  "facets": [
   "agent_readiness",
   "discoverability"
  ],
  "agent_readiness_dimensions": [
   "agentic_commerce"
  ],
  "why": "Agents as buyers break discovery-led retail; agentic commerce.",
  "abstract": "Agentic commerce is quietly breaking discovery-led retail. When an AI agent acts as an intermediary between the shopper and the store, the core mechanics of attribution, visual persuasion, and bidding auctions stop holding. Impressions are no longer served directly to human decision-makers, and persuasion strategies built for human psychology are priced into auctions that algorithms clear differently. In this session, Katia sets out what changes structurally for digital commerce and delivers an operational framework for what e-commerce, MarTech, and growth teams must instrument now to prepare for agent-mediated purchasing.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4861,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Two Architects, One Mission: Rethinking How Systems Talk",
  "speakers": [
   {
    "name": "Fabrizio Lazzaretti",
    "title": "Senior Managing Consultant",
    "company": "Wavestone"
   }
  ],
  "facets": [
   "contract_quality",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "Picking communication patterns (REST, events) by need, not habit.",
  "abstract": "Integration projects fail. Not because the technology doesn't work—it almost always does. They fail because communication patterns are picked based on habit rather than need. \"\"What did we use last time? REST? Great, let's do that again.\"\" These reflexes and other influences lead to the same issue: systems that may technically function but fundamentally misunderstand the business requirements. This leads to operational problems, incorrect outcomes, scaling issues, high costs, and migration efforts. We've spent years helping organizations untangle these messes across industries. The pattern is always the same: a mismatch between what the business needs from an interaction and what the technical implementation delivers. A synchronous call that would benefit from eventual consistency for resilience. An event stream where a simple request-response would suffice. Fire-and-forget where guaranteed delivery matters. In this session, we'll share what usually goes wrong when teams default to familiar patterns without examining requirements. We'll take you on a tour through the landscape of communication styles—synchronous and asynchronous, commands and events, orchestration and choreography—not as abstract theory but as tools with specific trade-offs. Most importantly, we'll give you a practical framework to diagnose what your current integration scenario actually demands. You'll leave with a decision model that starts with business intent rather than technology preference. Because the goal isn't to use Kafka everywhere—or REST everywhere—but to choose the pattern that makes your systems genuinely talk the way your business needs them to.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4662,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Signal Over Noise: Context Engineering as the Architecture Discipline Behind Agentic Enterprise AI",
  "speakers": [
   {
    "name": "Matthias Biehl",
    "title": "A(P)I Strategist",
    "company": "API-University.com"
   }
  ],
  "facets": [
   "agent_readiness",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Context engineering for enterprise agents; architecture, weak public-surface signal.",
  "abstract": "Your AI investments are delivering results but not consistently, and not at the scale you expected. The gap between the demo and production almost always comes down to the same root cause: the model doesn't know enough about your business at the moment it needs to act. Context engineering, the practice of designing what enterprise data an AI system has access to, when, and in what form is the most important and least understood discipline in enterprise AI today. The teams quietly building reliable AI systems are winning by getting the right enterprise data in front of the model at the right moment. In this session, we'll break down what context engineering actually is, why it's an architecture decision and not a prompting technique, and how to think about the two fundamental patterns, static injected context and dynamic live retrieval via tools like MCP. You'll leave with a clear mental model for diagnosing and building context for your enterprise that holds up in production.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4666,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Building Software in the AI Era",
  "speakers": [
   {
    "name": "Chouaieb Nemri",
    "title": "Generative AI Blackbelt",
    "company": "Google"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "agent_skills",
   "mcp_server"
  ],
  "why": "Agentic software engineering with Agent Skills and MCP.",
  "abstract": "Best practices to leverage Agentic Software Engineering, e.g. Antigravity, Gemini CLI and tools like (Agent Skills, MCP) to build software reliably faster.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4497,
  "day": "2026-12-03",
  "type": "Panel discussion",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Panel: API use inside DSI",
  "speakers": [
   {
    "name": "Emna Trigui",
    "title": "API designer & IT project leader",
    "company": "CEA"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "Panel on internal API use in IT departments and APIs for agents.",
  "abstract": "I wand to contribute at rond table our workshop session at french about differents topic around API use Case at DIS, API future in the IA Agent area.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4741,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:30",
  "end": "12:55",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "The Risks That Come with Underfunded API Standards",
  "speakers": [
   {
    "name": "Thulisile Sibanda",
    "title": "Senior Community Manager",
    "company": "AsyncAPI Initiative"
   }
  ],
  "facets": [
   "contract_governance",
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "Underfunded open API standards as a risk to shared infrastructure.",
  "abstract": "The technology industry is investing heavily in AI, accelerating software development and relying on increasingly interconnected systems. Beneath this rapid evolution lies shared digital infrastructure: the open standards, specifications, and APIs that enable software ecosystems to communicate, innovate, and scale. However, the resources required to sustain this critical infrastructure are failing to keep pace with industry adoption. When funding is uncertain, and only a few people remain to maintain these standards, the problem becomes more than financial; it becomes a security risk. Without enough support, API specifications go unfunded, security patches are delayed, and governance is overlooked. This makes software supply chains more vulnerable, especially as businesses rely on these systems more than ever. Drawing on real-world community insights from the AsyncAPI initiative and the broader open-source ecosystem, this session challenges the assumption that shared technical infrastructure will remain resilient simply because it is open. We will explore the friction between massive commercial adoption and maintainer capacity, examine the cybersecurity risks of neglected open specifications, and discuss actionable models for long-term sustainability, governance, and collective industry stewardship.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4949,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "12:40",
  "end": "13:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Sustainable Digital Value Chains: How Enterprise Architecture Enables IT Sustainability at Vodafone",
  "speakers": [
   {
    "name": "Rainer Scheibehenne",
    "title": "Enterprise Architect",
    "company": "ITM GmbH"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "IT sustainability via enterprise architecture at Vodafone; nothing the rubric reads.",
  "abstract": "As sustainability becomes a key differentiator in digital markets, IT sustainability at Vodafone is evolving from a compliance obligation to a strategic value driver. This session shows how Enterprise Architecture creates transparency across digital value chains and links IT assets to their environmental impact. We share insights from an ongoing transformation and demonstrate how EA supports data-driven sustainability management.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4577,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "12:40",
  "end": "13:00",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "The No Exit Dilemma: No Tech Fix Is Enough, No Bystanders",
  "speakers": [
   {
    "name": "Adel Noureddine",
    "title": "Full Professor of Computer Science",
    "company": "University Paris Nanterre"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Digital energy and environmental footprint; sustainability.",
  "abstract": "Our digital world has no off switch: modern societies depend on software and digital infrastructure. Yet its energy and environmental footprint keeps growing. That is the No Exit Dilemma. Unplugging is not an option, and technological efficiency is necessary but only buys us time. This talk explores what monitoring can achieve, what the numbers cannot tell us, and why human behavior is often the biggest lever we have. No tech fix is enough, and there are no bystanders.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4839,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:15",
  "track": [
   "FINOS"
  ],
  "room": [
   "Acacia 6"
  ],
  "title": "Introduction: Open source and FINOS",
  "speakers": [
   {
    "name": "Susan Haimet",
    "title": "Symphony",
    "company": "FINOS Ambassador"
   }
  ],
  "facets": [
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "FINOS intro: open source collaboration in finance.",
  "abstract": "Why the financial industry collaborates in the open, what FINOS does, and how the community is approaching AI.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4333,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Don't Blame the Agent. It Only Believed Your API Contract",
  "speakers": [
   {
    "name": "Deepti Mittal",
    "title": "Principal software engineer",
    "company": "Mimecast"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality",
   "operational_transparency"
  ],
  "agent_readiness_dimensions": [
   "error_semantics",
   "rate_limit_signal",
   "idempotency"
  ],
  "why": "Agents take the contract literally: error semantics, retries, unwritten assumptions.",
  "abstract": "Every REST API we've ever shipped carries assumptions it never wrote down. That the consumer read the documentation. That when something fails, it will eventually get tired and stop. That an error message is prose for a person. That \"try again later\" is a complete answer. That anyone holding a valid token is entitled to act on it. Those assumptions held for fifteen years, because on the other end was a developer with a browser tab open and a finite supply of patience. None of them survive an autonomous consumer. This is a code-first talk, not a tour of what agents are. This talks covers how to write discoverable capabilities instead of a README nobody opens, errors as control flow instead of apology text, mutations that survive being retried, an explicit contract for \"not yet,\" and an authority boundary for the actions that are too expensive to take unilaterally. Audience will leave with the five questions to ask in next design review — and a much clearer sense of whose fault the next incident actually is.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4346,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "An open business capability map for APIs and Enterprise Architecture",
  "speakers": [
   {
    "name": "Vincent Verdet",
    "title": "Senior Enterprise Architect",
    "company": "Independent"
   }
  ],
  "facets": [
   "discoverability",
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "operation_distinctiveness"
  ],
  "why": "Open business capability map organising API catalogues by capability.",
  "abstract": "A business capability map gives enterprise architects one shared language, from the board's heatmap down to the application inventory. It does the same for APIs : API catalogues are organised by supply: by team, gateway or endpoint. The business, and the AI agents acting for it, look by need. Yet most capability reference models are proprietary, and many EA functions still start from a blank page. This talk presents an open alternative: the Turbo EA capability catalogue, free under CC BY 4.0. It holds more than 9,600 capabilities across 27 industries, plus 1,300+ processes and 65 end-to-end value streams, in eight languages with stable IDs. Then the part I find most interesting: how it was built. The question was simple : could AI build a capability model from the ground up ? From defining the governance model to validating a multi-step approach, I will show how to use a non-deterministic model to produce near-deterministic output, industry after industry.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4353,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Building usage-based billing for banks",
  "speakers": [
   {
    "name": "Hassane Moustapha",
    "title": "Architect and Coach IT",
    "company": ""
   }
  ],
  "facets": [
   "access_clarity",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "Usage-based billing and pricing logic for bank APIs.",
  "abstract": "Banks already charge for everything: transfers, card payments, statements, API calls. The pricing logic usually lives in spreadsheets, in core banking parameters, and in people's heads. When a regulator changes a rule, it takes months to follow. My team and I built a platform to fix this. It's a usage-based billing platform: every event comes in through an API, gets metered, gets rated against a pricing plan, and ends up on an invoice. It's designed for banks and their subsidiaries. In this talk I'll walk through the real architecture and the problems that come with it: idempotent ingestion keeping pricing calculations consistent across services re-billing from a group to its subsidiaries adapting to regulatory pricing changes without a rewrite.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4596,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "LVMH × Sopht : du Green IT mesuré au Green IT piloté",
  "speakers": [
   {
    "name": "Damien Nierat",
    "title": "Head of IT End User Experience & Green IT",
    "company": "Christian Dior (LVMH)"
   },
   {
    "name": "Théo Blechschmidt",
    "title": "Delivery & Customer Success Manager",
    "company": "Sopht"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Measuring and steering IT footprint at LVMH; sustainability.",
  "abstract": "Chez LVMH, la mesure de l’empreinte IT couvre plus de 20 Maisons du groupe, avec une collecte quasi automatisée. Damien Nierat (Christian Dior) et Théo Blechschmidt (Sopht) racontent comment le groupe est passé de la mesure au pilotage vers 2030. Ils abordent ensuite les prochains chantiers : l’empreinte de l’IA et celle du SaaS, en carbone comme en euros.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4328,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Beyond the Anti-Corruption Layer: The use of APIs to efficiently leverage, manage, and protect Agents",
  "speakers": [
   {
    "name": "Radhouane Jrad",
    "title": "Enterprise Integration Architect",
    "company": "DLA Piper"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance"
  ],
  "agent_readiness_dimensions": [],
  "why": "Using APIs to leverage, manage and protect agents.",
  "abstract": "Agentic AI is a hot topic, and APIs are a mature concept. Using one or the other returns a lower business value than when combining both.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4736,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Your AI Agent Is Only as Good as Its Schema: Designing Reliable MCP Tools with JSON Schema",
  "speakers": [
   {
    "name": "Ayodele Aransiola",
    "title": "Solution Architect/Developer Advocate",
    "company": "Voyatek Group"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_tool_quality",
   "mcp_server",
   "spec_presence"
  ],
  "why": "JSON Schema design for reliable MCP tool calls.",
  "abstract": "AI agents are becoming active consumers of APIs, but many of the contracts they rely on were designed only for traditional applications and human developers. That creates ambiguity, unreliable tool calls, inconsistent outputs, and unnecessary complexity when teams introduce MCP or other agent-based systems. This talk explores why JSON Schema is becoming critical infrastructure for reliable AI tooling. Using MCP tools as a practical example, I’ll show how well-designed schemas can make agent interactions more predictable, improve validation, reduce ambiguity, and create clearer contracts between models and backend systems. The goal is to move the conversation beyond prompts and model choice, and focus on something more fundamental: if an agent cannot clearly understand what a tool accepts and returns, the quality of the model alone will not make the system reliable.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4412,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "AI - Spec-Driven Development: Lessons from £Billions in Payments",
  "speakers": [
   {
    "name": "Cihan Dogan",
    "title": "Senior Software Engineer",
    "company": "Bally's Intralot"
   }
  ],
  "facets": [
   "contract_quality",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Spec-driven development on a live payments platform.",
  "abstract": "Letting AI write code is one thing. Letting it write code that moves real money is another. This is not a tool demo. It is a year of spec-driven development on a live payment platform that processes billions of pounds in deposits and withdrawals, across seven real projects that are all running in production today. AI can write a week of code in an afternoon, but most teams still ship at the same speed they always did. Spec-driven development is what closed that gap for us: work that used to take months now takes weeks, on a PCI-regulated payment platform serving 39+ countries. In our payment team we changed the order of work. We agree the spec first, then let AI write the code. In this talk I walk through two real integrations shipped this way, a secondary provider for Apple Pay and Google Pay built with GitHub Spec Kit, and a new payment method integration delivered in five weeks before the World Cup deadline, with the production numbers behind both. I will also cover what it takes to make this work at scale: keeping AI focused in a large codebase, where the process slows you down instead of speeding you up, and the honest pros and cons of running this on a system where a mistake costs real money.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4554,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Peut-on allier industrialisation et IA frugale ? Le platform engineering au service de la maîtrise de nos usages IA",
  "speakers": [
   {
    "name": "Benjamin Demonet",
    "title": "Product Owner & Référent Numérique Responsable",
    "company": "SNCF Connect & Tech"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Platform engineering for frugal AI use; sustainability.",
  "abstract": "L'IA générative change brutalement l'échelle du problème : modèles puissants, usages qui se multiplient, boucles agentiques dont on ne perçoit plus les impacts. Sans maîtrise, elle peut effacer des années d'écoconception. Mais industrialiser ne rend pas l'IA frugale : sans contraintes, on provoque un effet rebond. Retour d'expérience de SNCF Connect & Tech (labellisée Numérique Responsable niveau 2) : industrialiser pour mesurer et arbitrer, mutualiser l'agentique comme un produit de plateforme, et préserver notre capacité à fonctionner sans l'IA.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4501,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:00",
  "end": "14:25",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Making official standards AI friendly",
  "speakers": [
   {
    "name": "Frank Terpstra",
    "title": "",
    "company": "Geonovum"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; making official standards AI friendly.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4571,
  "day": "2026-12-03",
  "type": "Panel discussion",
  "start": "14:00",
  "end": "14:40",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Green IT in Europe: How policy, transparency and procurement can shape a sustainable Tech future",
  "speakers": [
   {
    "name": "Wiebren van der Zee",
    "title": "Domain Expert Sustainable IT, CIO Office",
    "company": "ABN AMRO Bank"
   },
   {
    "name": "Pieke Hezemans",
    "title": "Project Advisor Public Affairs & Online Communication",
    "company": "Dutch National Coalition for Sustainable Digitalisation"
   },
   {
    "name": "Ekaterina Smid-Gankin",
    "title": "Sr. Circular Procurement Advisor",
    "company": "Circular and Fair ICT (CFIT) Pact Secretariat, Ministry of Economic Affairs and Climate Policy"
   },
   {
    "name": "Analivia Brouwer",
    "title": "Policy Coordinator Sustainable AI, European & International Twin Transition Policy",
    "company": "Directorate of the Digital Economy, Ministry of Economic Affairs and Climate Policy"
   }
  ],
  "facets": [
   "outside",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "EU policy, transparency and procurement for green tech; sustainability.",
  "abstract": "AI and digital tech draw growing attention for their energy use, water use, data centres and demand for critical raw materials. Europe's dependence on external providers and global supply chains limits transparency and control. This panel brings three perspectives together: European policy direction, transparency and measurement (including the proposed GenAI Sustainability Rating), and public procurement as a lever for market change. The panel concludes on how policy, measurement and procurement can work together to turn transparency into better decisions and real market change.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4676,
  "day": "2026-12-03",
  "type": "Panel discussion",
  "start": "14:00",
  "end": "15:25",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Panel: Nobody read this code. Ship it? AI-generated code, human review, and who answers for the release",
  "speakers": [
   {
    "name": "Pier-Jean Malindrino",
    "title": "CTO",
    "company": "Scub"
   },
   {
    "name": "Kapil Bakshi",
    "title": "CEO/Co-founder",
    "company": "Karate Labs"
   }
  ],
  "facets": [
   "accountability",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; accountability for AI-generated code releases, not API surface.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4840,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:15",
  "end": "14:40",
  "track": [
   "FINOS"
  ],
  "room": [
   "Acacia 6"
  ],
  "title": "AI governance in financial services",
  "speakers": [
   {
    "name": "Elena Guz",
    "title": "Financial Risk Manager",
    "company": "Morgan Stanley"
   }
  ],
  "facets": [
   "regulatory",
   "accountability"
  ],
  "agent_readiness_dimensions": [],
  "why": "AI risk governance and evidence for financial regulators.",
  "abstract": "A risk manager's view on identifying AI risks, setting governance and accountability, and evidencing it to regulators and auditors.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4555,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:25",
  "end": "14:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Qualité et traçabilité des données : comment réduire l’empreinte de l’IA dès la source",
  "speakers": [
   {
    "name": "Jawaher Allala",
    "title": "CEO",
    "company": "Systnaps"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Data quality and traceability to reduce AI footprint.",
  "abstract": "Réduire l’empreinte de l’IA commence par une meilleure confiance dans les données et une réflexion sur leur parcours. Comment limiter la collecte, le stockage et les traitements inutiles ? Quand un algorithme déterministe suffit-il, et quand l’IA apporte-t-elle une réelle valeur ? Cette intervention propose une démarche concrète, de la donnée à l’algorithme puis à l’IA, pour choisir la solution la plus sobre sans perdre en fiabilité.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4936,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "14:25",
  "end": "14:50",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Extending IT hardware life: Evernex's circular IT playbook",
  "speakers": [],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Hardware life extension and e-waste; sustainability.",
  "abstract": "Every server, storage array and switch kept in service is e-waste avoided and emissions not spent on new hardware. Evernex, a global IT Life Services provider with 40 years of experience in asset life extension, shares how third-party maintenance, refurbishment and responsible IT asset disposition keep data-centre equipment running well beyond the manufacturer's end of service life. From its Renew Center near Paris, EMEA's largest reconditioning and recycling facility for second-hand IT hardware, the workshop shows how to measure and scale circular IT.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4332,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Connecting the Dots  -  From scattered dots to one autonomous flow — Idea to Production",
  "speakers": [
   {
    "name": "Jason Feng",
    "title": "Principal Software Engineer / Consultant",
    "company": "Mantel"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "agent_skills",
   "mcp_server"
  ],
  "why": "Stitching coding agents, skills, MCP and evals into an SDLC flow.",
  "abstract": "Everyone in the AI world already knows the dots: coding agents, skills, sub-agents, MCP, evals, hooks. What almost nobody has is the line through them — most teams run brilliant fragments while their SDLC stays stubbornly manual in between. ▎ This talk shows the line. I'll walk through how we turned a plugin marketplace repo into a workflow factory: agents that do the work, deterministic gates that block bad output, human approval as the only manual labour, and accepted specs as the system's memory. Then I'll run it — one command that takes a rough idea through research, requirements, test cases, design, and plan, pauses twice for a human \"\"yes\"\", and ends by writing, testing, and committing the code itself: a review-ready branch, every step gated on the project's own definition of green. Not a concept demo — this pipeline shipped its first live feature in 21 minutes, and its telemetry ledger is how the remaining approvals will earn their way out. ▎ You'll leave with the architecture, the guardrails that make autonomy safe to grant, and a repo pattern your team can copy on Monday.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4863,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Designing the Sovereign Agent Stack: MCP, A2A, and API Governance for the Agentic Era",
  "speakers": [
   {
    "name": "Jon Scheele",
    "title": "Founder & Principal Consultant",
    "company": "Blue Connector"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_governance",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "agent_card",
   "delegated_identity"
  ],
  "why": "Governing agent discovery and action via MCP and A2A with sovereignty.",
  "abstract": "Agentic AI raises a sovereignty problem: how do you let autonomous agents discover and act on your systems without losing control over data residency, auditability, and vendor lock-in? MCP and A2A give architects the plumbing, but not the governance model. This session maps the emerging protocol stack against the EU AI Act, NIST's AI Risk Management Framework, and Singapore's MAS model, translating three regulatory approaches into one practical architecture for building agent systems that are sovereign and governed by design, not by afterthought. Key takeaways: • Where MCP and A2A each fit in a sovereign agent stack, and why most enterprises need both • How EU AI Act, NIST AI RMF, and MAS requirements compare — and where they converge • Governance patterns for agent-to-agent coordination that preserve auditability and data residency",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4500,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "European Commission's API Framework for Digital Governments: 5 Years Later",
  "speakers": [
   {
    "name": "Mark Boyd",
    "title": "Director",
    "company": "Platformable"
   }
  ],
  "facets": [
   "contract_governance",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; EC API framework for digital governments, five years on.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4419,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Your API Has a New User. It Doesn’t Read the Docs, It Has 200,000 Tokens of Attention, and It Will Never Call Support.",
  "speakers": [
   {
    "name": "Mehdi Yadir Goldfeil",
    "title": "Fractional CPO/CTPO",
    "company": "My Super Academia"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [
   "error_semantics",
   "mcp_tool_quality",
   "operation_distinctiveness"
  ],
  "why": "APIs built for humans fail agents: hallucinated params, 400 loops, wrong endpoints.",
  "abstract": "MCP is just an API,” they say. True, and that is exactly the problem. We hand AI agents APIs designed for humans, then act surprised when they hallucinate parameters, loop on 400 errors or call the wrong endpoint. Your API has a new consumer. It never reads your docs, it runs on a fixed attention budget called a context window, and it will never open a support ticket. It simply fails, retries, or quietly works around you, just like the employee who goes back to their trusty spreadsheet. The good news: we already know how learners succeed. This talk maps neuroscientist Stanislas Dehaene’s four pillars of learning (attention, active engagement, error feedback and consolidation) onto the design of APIs consumed by agents. You will see why a tool description is a prompt, why an error message is a lesson, why eighty CRUD endpoints are worth less than six intent-based operations, and how a consistent contract lowers cognitive cost for machines and humans alike. Expect real before-and-after examples, a few lessons borrowed from military strategy, and a twelve-question Agent-Readiness checklist you can run on your own APIs on Monday morning. Because we no longer design APIs for readers. We design them for learners.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4641,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Engineering Multiplayer API Sandboxes for Humans and Agents",
  "speakers": [
   {
    "name": "Nithin Kumar B",
    "title": "Founding Engineer",
    "company": "Wundergraph"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "dry_run_mode"
  ],
  "why": "Multiplayer API sandboxes keeping schema edits valid for humans and agents.",
  "abstract": "Many multiplayer apps sync edits with a CRDT, a data structure that merges changes. That works for text or most design tools. An API schema, be it OpenAPI, gRPC or GraphQL, has rules, and two edits that are each fine can break it together. We learned this building API sandboxes, private copies of an API that people and AI agents edit together, on schemas with over 200 services. So we built our own sync engine. Every edit, from a person or an agent, gets a clear answer on whether it worked. Large schemas also outgrew a regular web canvas, so we built one that draws on the GPU. The canvas is where people follow an agent at work. It brings changes from agents into view, and describes itself to agents in return. You don’t need to know CRDTs or graphics to follow along. I’ll share why we stopped using the usual tools, and what carries over to your APIs.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4337,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:30",
  "end": "14:55",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "Your API changed, did your docs notice?",
  "speakers": [
   {
    "name": "Kim Rowden",
    "title": "Founder",
    "company": "LaunchDocs"
   }
  ],
  "facets": [
   "developer_ergonomics",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Keeping docs in sync as APIs change fast.",
  "abstract": "AI is dramatically increasing the speed at which software and APIs are built, changed, and shipped. Documentation, however, is still largely maintained through manual processes. That creates an increasingly important gap: the product evolves continuously, while the documentation describing it struggles to keep up. The deeper issue is that documentation is often treated as something that happens after development rather than as part of the development lifecycle itself. As AI changes how software is built, we have an opportunity to rethink this model. Documentation can become aware of changes across code, tickets, specifications, and engineering context, identify what has become outdated, and propose updates while keeping humans in control. The result is documentation that evolves alongside the API rather than constantly chasing it.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4742,
  "day": "2026-12-03",
  "type": "Panel discussion",
  "start": "14:30",
  "end": "15:25",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "AsyncAPI Ambassador Panel - Can AsyncAPI help AI Agents Speak Event-Driven?",
  "speakers": [
   {
    "name": "Hugo Guerrero",
    "title": "Developer Advocate",
    "company": ""
   },
   {
    "name": "Naresh Jain",
    "title": "Founder & CEO",
    "company": "Specmatic"
   },
   {
    "name": "Laurent Broudoux",
    "title": "Co-founder",
    "company": "Microcks & Reshapr"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "event_surface_described"
  ],
  "why": "No abstract; AsyncAPI for agents in event-driven systems.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4579,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:40",
  "end": "15:05",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Your Carbon Dashboard Is Not Saving the Planet",
  "speakers": [
   {
    "name": "Wilco Burggraaf",
    "title": "Sustainable Digital Architect",
    "company": "HighTech Innovators"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Carbon dashboards versus actual reduction; sustainability.",
  "abstract": "Carbon dashboards make emissions visible, but visibility isn't reduction—it doesn't retire idle workloads or simplify architecture. Telling engineers to build greener software isn't enough. What if sustainability was baked into how we govern, design, and build—rather than measured afterwards? In this 20-minute talk, I challenge Green IT’s carbon-first reflex and explore governing resource demand. How can smart defaults and policies nudge teams toward frugality without slowing innovation? Building a sustainable digital Europe requires making efficiency an everyday engineering standard. The goal: turn measurement into action, cut digital waste, and make lower emissions a natural consequence of better engineering. Followed by 5 minutes of Q&A.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4841,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:40",
  "end": "15:20",
  "track": [
   "FINOS"
  ],
  "room": [
   "Acacia 6"
  ],
  "title": "Defining controls and policies with CALM",
  "speakers": [
   {
    "name": "James Gough",
    "title": "Distinguished Engineer",
    "company": "Morgan Stanley"
   }
  ],
  "facets": [
   "contract_governance",
   "regulatory",
   "open_source"
  ],
  "agent_readiness_dimensions": [],
  "why": "FINOS CALM: architecture, controls and policies as code.",
  "abstract": "Using FINOS CALM to express architecture, controls and policies as code and enforce them across the delivery lifecycle.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4549,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "14:50",
  "end": "15:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Sustainable Digital Challenge : la finale et la remise des prix",
  "speakers": [
   {
    "name": "Nicolas Drouet",
    "title": "CEO",
    "company": "The Green Compagnon"
   },
   {
    "name": "Romain Bachelet",
    "title": "Référent Numérique Responsable",
    "company": "SNCF Voyageurs Matériel"
   },
   {
    "name": "Sandra Sydow",
    "title": "President & Cofounder",
    "company": "Planet On Stage (NGO)"
   },
   {
    "name": "Amine Saboni",
    "title": "ML Engineer",
    "company": "Code Carbon"
   },
   {
    "name": "Ludi Akué",
    "title": "Chief Technology Officer",
    "company": "Bpifrance"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Award ceremony, no API relevance.",
  "abstract": "Annonce des lauréats du Sustainable Digital Challenge et remise des prix, avec photo des équipes. Organisé par The Green Compagnon.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4347,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "From Concrete to Code : Bringing Modern APIs to the Energy Giants",
  "speakers": [
   {
    "name": "Aya Lachbi",
    "title": "Cybersecurity Support, ENGIE Supply & Energy Management B2B",
    "company": "Engie"
   }
  ],
  "facets": [
   "discoverability",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [],
  "why": "Bringing API-first design to the energy sector.",
  "abstract": "The energy sector is historically built on concrete, steel, and decades of rigorous safety standards—not necessarily agile sprints or API-first design. For a long time, \"engineering\" meant building physical infrastructure. But as the energy transition accelerates, that definition is shifting rapidly. In this session, I share my on-the-ground experience as an apprentice at Tractebel (ENGIE Group), working at the intersection of heavy industry and modern software architecture. This talk explores the unique challenge of deploying modern API strategies within a legacy industrial environment. I will discuss how we break down data silos between traditional engineering teams and IT, and the cultural friction that comes with replacing manual workflows with automated connectivity Because the future of energy isn't just about atoms; it’s about bits. As traditional industries digitize, the ability to wrap complex physical systems in clean, consumable APIs is becoming a critical skill. Whether you are a senior architect or a fellow junior, you will gain a fresh perspective on how \"real-world\" industries are entering the API economy, and why the most impactful code today is often written where you least expect it.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4356,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "The Ocean is a Product: Building Sovereign APIs for the Blue Economy",
  "speakers": [
   {
    "name": "Cellou Diallo",
    "title": "Digital Ocean Architect",
    "company": "Blue Connex"
   }
  ],
  "facets": [
   "discoverability",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "Making ocean data accessible via sovereign APIs.",
  "abstract": "It is often said that we know more about the surface of the Moon than the depths of our own oceans. However, the real challenge today isn't just data collection—it's accessibility. Between satellite feeds, acoustic sensors, and subsea drones, the BlueTech sector is drowning in data, yet remains trapped in proprietary silos and 20th-century Excel files. ​In this session, we will explore how to apply the API-as-a-Product methodology to one of the most complex environments on Earth. I will share my 'battle scars' and insights from building Blue Connex, a sovereign API infrastructure designed to transform maritime 'Data Swamps' into seamless, secure, and interoperable services. ​We will discuss how to reconcile two often-conflicting imperatives: radical openness (for scientific innovation) and strict sovereignty (for the protection of strategic national resources). No hype—just practical lessons on designing interface contracts (OpenAPI) for the high seas, managing latency in offshore environments, and building trust-based gateways for the future of the Blue Economy.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4325,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "The Non-Human Developer: Securing APIs for AI Agents in a Post-DORA World",
  "speakers": [
   {
    "name": "Harry Bouras",
    "title": "Head of API",
    "company": "Colonnade"
   }
  ],
  "facets": [
   "agent_readiness",
   "access_clarity",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "rate_limit_signal",
   "auth_clarity",
   "agent_identity_declared"
  ],
  "why": "Securing APIs against agent traffic spikes under DORA.",
  "abstract": "For the past decade, we optimized API Developer Experience (DevEx) for humans. Today, our fastest-growing API consumers are autonomous AI agents. Unlike humans, agents are unpredictable, can trigger massive traffic spikes via \"hallucination loops,\" and consume millions of tokens. In highly regulated industries, an un-governed AI agent isn't just an IT headache under strict DORA enforcement, it is a severe compliance risk. This session explores the transition from DevEx to \"Agent Experience\" (AX) and how to architect Zero-Trust API gateways to protect enterprise core systems. Key Takeaways & Outline: To keep it engaging, I will anchor the talk in real-world metrics and architectural blueprints: The Paradigm Shift: Why agentic workflows have officially surpassed human-driven API consumption. The Chaos of the Machine: Quantifying the \"Hallucination Loop\" (e.g., how a confused AI agent can accidentally trigger 36k requests/hour) and massive LLM token costs. The Regulatory Collision: The DORA reality, how an un-governed AI causing API downtime can lead to penalties up to 2% of global turnover. Architecting the \"Agent-Proof\" API: Practical blueprints for semantic rate-limiting, dynamic PII masking at the edge, and intent-based verification.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4698,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Europe API Conference"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "Closing Note: The Future of Public Services APIs",
  "speakers": [
   {
    "name": "Christophe Gaie",
    "title": "Directeur adjoint du Pôle Opérations",
    "company": "DSN"
   }
  ],
  "facets": [
   "contract_governance",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; closing note on public-services APIs.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4392,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Jev as a concept and how deterministic systems are the answer to Financial Services and similar industries",
  "speakers": [
   {
    "name": "Deepak Pandiyarajan",
    "title": "Vice President",
    "company": "JP Morgan"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Deterministic systems over LLMs in financial services; no API surface stated.",
  "abstract": "Whenever we think of AI, we think of it as a Chatbot and everything we do is prone to hallucinations and exponential token usage. We seem to be at the mercy of an unknown beast. Its time we take control of what we want to build and how we want to build by using a more deterministic system. Although my proof of concept is with lightweight payments and code refactoring from an existing knowledge base, the concept of deterministic flows will need to be embedded into the fabric of the organization than commanding a chat bot. I will give a real life practice use case, a demo and some hands on guidance for participants.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4667,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:00",
  "end": "15:25",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "Hallucinating in Three Dimensions: Lessons Learned from Trying Every 3D GenAI Tool",
  "speakers": [
   {
    "name": "Adèle Keane",
    "title": "Creative Director",
    "company": "Adèle Keane"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "3D generative AI tool lessons; no API relevance.",
  "abstract": "When a text LLM hallucinates, it fabricates a fact. When a 3D model hallucinates, it breaks physics, inverts normals, and crashes pipelines. In immersive media, 3D animation, and game design, that's a real showstopper. As studios rush to adopt generative AI across film, animation, XR, and virtual production, errors like impossible topology and asset inconsistencies are creeping into projects, often only discovered after hours have been wasted trying to integrate them into Unity or Unreal. In this talk, I'll share lessons learned from experimenting with a wide range of 3D GenAI tools, what works, what doesn't, and some practical tips and best practices for using generative 3D media in creative pipelines.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4872,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:05",
  "end": "15:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Data Centre Essentials - Sustainability in DCs",
  "speakers": [],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Data centre sustainability book; nothing the rubric reads.",
  "abstract": "The second edition of the book Data Centre Essentials: Design, Construction, and Operation of Data Centres for the Non-expert by Vincent Fogarty and Sophia Flucker was launched in November. The book provides a good overview of all things DC. Part of it is sustainability. This talk will focus on Chapter 9 - Sustainability and provide an overview of the sustainability landscape in data centres (regulations, considerations, methodology) and important aspects of AI and its impact.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4932,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "15:10",
  "end": "15:30",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "TokenOps: Trends, Metrics and Best Practices for Navigating the Energy Constraints, Costs and Impacts of Agentic AI",
  "speakers": [
   {
    "name": "Ismael Velasco",
    "title": "CEO",
    "company": "Adora Foundation"
   }
  ],
  "facets": [
   "outside",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "Token cost and energy of agentic AI; only finops brushes Access Clarity.",
  "abstract": "This workshop introduces trends, metrics and tools for more frugal, effective and sustainable AI: multi-model orchestration, effort tuning, prompt and harness optimisation, cache-first design and systematic evaluation. It demos tools that model planetary energy constraints and agentic consumption trends, pointing to rising compute scarcity and costs, and a custom instrumentation of OpenAI's Codex harness tracking token intensity turn by turn, session by session and model by model. It ends with tooling for a full lifecycle assessment of the non-AI compute behind our digital products.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4782,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "15:55",
  "end": "16:15",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "Measuring the impact of coding agents with agentic-footprint",
  "speakers": [
   {
    "name": "Amine Saboni",
    "title": "ML Engineer",
    "company": "Code Carbon"
   },
   {
    "name": "Benoit Courty",
    "title": "President",
    "company": "CodeCarbon"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Measuring coding-agent footprint; sustainability.",
  "abstract": "The recent development of agentic based systems have grown the LLM usage by orders of magnitude. This growth has been measurable at the macro level with the deployment of datacenters full of GPUs, but at the local level, developers have few options to understand the impact of using tools such as Claude Code or other coding agents. During this workshop we will see how the agentic-footprint package combines the estimation of remote APIs using EcoLogits, with CodeCarbon to measure the consumption of local runtime of those agents and monitor those local use cases with popular coding agents.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4646,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 4"
  ],
  "title": "Making sense of the application Frankenstack in AI times",
  "speakers": [
   {
    "name": "Paolo Malinverno",
    "title": "Chief strategy officer",
    "company": "Sensedia"
   }
  ],
  "facets": [
   "contract_governance",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "mcp_server"
  ],
  "why": "Rationalising API, event, MCP and AI gateway management layers.",
  "abstract": "how the various Frankenstack components came along (API management, Events management; MCP management; LLM management, Context Management, but also identity, MCP servers and AI gateway) just dumping all AI stuff on top of that mess won't work if you think designing AI agents on top of that is complicated, wait to see how AI Agents (they are new users) proliferation will make unprecedented demands on it how to make sense of the mess: the history is repeating itself, just enough governance sadly is going to prove an illusion yet again a few myths to get out of the way (e.g. you don't need to do as many agents as possible, or single X is a myth, multi X is a reality) Model routing, MCP gateways, Agent gateways bring a lot of promises You don't have to use AI for everything, because it doesn't do everything: use human and AI agent teams, design AI Agents right from the beginning and protect your Frankenstack",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4842,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "FINOS"
  ],
  "room": [
   "Acacia 6"
  ],
  "title": "Scheduling AI workloads with High Performance Computing",
  "speakers": [
   {
    "name": "Raphael Javaux",
    "title": "Senior Engineer",
    "company": "Citi"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "HPC scheduling of AI workloads; no API surface.",
  "abstract": "How HPC and workload scheduling help organisations prioritise, run and scale AI jobs efficiently and cost-effectively.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4947,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 3"
  ],
  "title": "API Design: Is Your API Making Your Agent Hallucinate?",
  "speakers": [
   {
    "name": "Ketty Zakariasy",
    "title": "Co-Founder",
    "company": "So-API"
   },
   {
    "name": "Stephane Tuot",
    "title": "Co-founder",
    "company": "So-API"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "operation_distinctiveness",
   "openapi_examples"
  ],
  "why": "Same agent, three API designs: contract shape drives hallucination and tokens.",
  "abstract": "Live demo on an investment banking P&L use case. Same AI agent, same business question, 3 API designs. We change the API contract and compare what the agent makes up and how many tokens it uses.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4589,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Piloting EU-Hosted AI for EV Charging Customer Service",
  "speakers": [
   {
    "name": "Florence Bellendir",
    "title": "Enterprise Process and AI Innovation Lead",
    "company": "Mer"
   }
  ],
  "facets": [
   "outside",
   "regulatory"
  ],
  "agent_readiness_dimensions": [],
  "why": "EU-hosted AI pilot for EV charging customer service.",
  "abstract": "AI promises efficiency, scalability and better customer experiences. But what happens when you put those promises to the test? Mer’s Lemonflow pilot explores EU-hosted AI in real-world EV charging customer service and operations. This talk shares what worked, what didn’t, and what we learned about performance, data, governance, human oversight, ROI and CO₂ impact. It also explores the operational changes needed to succeed and introduces our next test: moving from reactive support to AI-powered proactive monitoring that can identify charging issues before drivers report them.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4859,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "GENIA - Testing - Industrialiser la QA logicielle avec une architecture Multi-Agents.",
  "speakers": [
   {
    "name": "Betty Beauge",
    "title": "RESPONSABLE TESTING FACTORY",
    "company": "NIJI"
   },
   {
    "name": "Gilles Nouais",
    "title": "Test Manager et coach IA QA",
    "company": "Niji"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Multi-agent QA for generated code; not API surface.",
  "abstract": "Avec l'IA générative, les développeurs codent plus vite. Le corollaire ? Le risque de générer de la dette technique et des anomalies à grande vitesse. Comment l'Assurance Qualité (QA) peut-elle tenir ce rythme industriel sans sacrifier la rigueur des standards ? Cette conférence présente notre retour d'expérience terrain au sein d'une Testing Factory, à travers \"\"QA Forge\"\" : un framework DevSecOps où l'IA a été industrialisée. Je vous présenterai comment nous sommes passés du simple \"\"prompt\"\" à une architecture orchestrant 17 agents IA spécialisés sur la qualité. De l'analyse \"\"Shift-Left\"\" d'une User Story Jira (critères INVEST, risques) à la génération de code automatisé (Playwright pour l'UI, Bruno pour les API), l'IA intervient à chaque étape. Mais le véritable enjeu abordé sera la confiance : comment éviter les hallucinations ? Nous détaillerons notre approche \"\"Supervised LLM\"\" (où une IA évaluatrice audite le travail de l'IA productrice) et notre infrastructure \"\"Agents as Code\"\" utilisant Cursor, le protocole MCP et Git comme source de vérité. Cette présentation démontre concrètement comment l'IA ne remplace pas la QA, mais la transforme. Le testeur devient un \"\"Testeur Augmenté\"\", architecte et garant d'une qualité logicielle générée \"\"by design\"\".",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4946,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Why MCP Needs an OpenAPI Moment",
  "speakers": [
   {
    "name": "Stève Sfartz",
    "title": "Principal Architect",
    "company": "Cisco"
   }
  ],
  "facets": [
   "agent_readiness",
   "contract_quality"
  ],
  "agent_readiness_dimensions": [
   "mcp_server",
   "mcp_tool_quality",
   "spec_presence"
  ],
  "why": "MCP needs a reviewable description format the way APIs got OpenAPI.",
  "abstract": "The MCP protocol standardizes how clients and servers communicate at runtime. But how do we reason about an MCP server before we implement it, deploy it, or connect an agent to it? The API ecosystem went through a similar transition. OpenAPI made APIs reviewable, documentable, testable, lintable, and governable before runtime. MCP now has an opportunity to establish a comparable design-time layer. This talk explores what a machine-readable MCP server description could unlock: design-first workflows, generated documentation, quality rules, contract reviews, change detection, code generation, and governance. We will use mcpdesc, an open specification for describing MCP server interfaces, published through the mcpdesc.org initiative, as a concrete exploration of what such a design-time layer could look like.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4948,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 6"
  ],
  "title": "The Anatomy of an Agentic AI Sandbox",
  "speakers": [
   {
    "name": "Ludovic Pourrat",
    "title": "API Architect & API Catalyst",
    "company": "Lombard Odier Group"
   }
  ],
  "facets": [
   "agent_readiness",
   "developer_ergonomics",
   "regulatory"
  ],
  "agent_readiness_dimensions": [
   "dry_run_mode",
   "agentic_access"
  ],
  "why": "Sandbox giving agents safe access to bank systems.",
  "abstract": "Agents are only as useful as the systems they can reach. In a regulated bank, those systems are the hardest ones to open up. So most agentic ideas stall before anyone can find out whether they work. At Lombard Odier, we took a different approach. We built a sandbox that gives agents everything they need and nothing they can break: realistic APIs, synthetic data, discoverable tools and documentation written for machines as well as people. Teams go from an idea to a working, validated agentic use case in days, without writing code. When failing is cheap and safe, teams try more ideas, faster. This talk walks through how the sandbox is built, and where we go next.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4370,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "Apidays"
  ],
  "room": [
   "Baobab 3"
  ],
  "title": "Notes from a Tech Futurist: Mapping the next five years in software technologies",
  "speakers": [
   {
    "name": "Mark Boyd",
    "title": "Director",
    "company": "Platformable"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Futurist overview of tech eras; no specific API topic.",
  "abstract": "Globally distributed digital architecture, APIs and the cloud already transformed society around the world. Now AI is upending it again. Next comes quantum computing. But not everyone has shifted to the new lanes, so we have three different technological eras co-existing. This presentation brings together a wide range of themes and trends to situate attendees of FOST in understanding the overall context that will influence their career over the next 1-5 years and their FOST experience for the next few days.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4550,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:20",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Do you really need an LLM for that? How Orange rethinks AI efficiency.",
  "speakers": [
   {
    "name": "Romain Carbou",
    "title": "Data Scientist for CSR",
    "company": "Orange"
   },
   {
    "name": "Rose Djagbre",
    "title": "Responsable Data Empreinte Carbone",
    "company": "Orange"
   }
  ],
  "facets": [
   "outside",
   "access_clarity"
  ],
  "agent_readiness_dimensions": [],
  "why": "LLM proxy to cut token cost and footprint.",
  "abstract": "Token consumption by apps is skyrocketing, leading to significant costs for both the planet and your margins. Sometimes, the answer to \"Do you really need an LLM for that?\" is simply no. That’s why we built a smart LLM proxy: it goes beyond mere monitoring, observes LLM calls without interfering, uncovers recurring patterns that could be handled by lighter approaches, and maintains a library of resource-efficient ML pipelines. It then recommends these alternatives and conducts live A/B tests, comparing quality, cost, lifecycle impact against the LLM. With an open-source mindset in mind.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4677,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:40",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Live coding: fixing a broken test suite, from contract to E2E",
  "speakers": [
   {
    "name": "Peter Thomas",
    "title": "Co-Founder & CTO",
    "company": "Karate Labs"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics"
  ],
  "agent_readiness_dimensions": [],
  "why": "No abstract; fixing tests from contract to E2E.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4950,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "15:55",
  "end": "16:50",
  "track": [
   "API Standards Conference"
  ],
  "room": [
   "Aubépine 5"
  ],
  "title": "Unconference: API standards 2027 working group",
  "speakers": [
   {
    "name": "Thulisile Sibanda",
    "title": "Senior Community Manager",
    "company": "AsyncAPI Initiative"
   },
   {
    "name": "Erik Wilde",
    "title": "Chief Evangelist",
    "company": "WunderGraph"
   },
   {
    "name": "Martin Bonnin",
    "title": "Mobile Engineer",
    "company": "Apollo GraphQL"
   },
   {
    "name": "Onyedikachi Hope Amaechi-Okorie",
    "title": "Technical Community Advocate",
    "company": "JSON Schema"
   }
  ],
  "facets": [
   "contract_governance"
  ],
  "agent_readiness_dimensions": [],
  "why": "Unconference for an API standards working group.",
  "abstract": "\"Unconference\" starts with U! Do you have a demo to share, an itch to scratch, ideas about where to take the conference? There's ample opportunity to bring your thoughts to the unconference table and seek or share feedback. The unconference agenda will be created onsite - stay tuned for more info about how you can add your topics.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4597,
  "day": "2026-12-03",
  "type": "Workshop",
  "start": "16:15",
  "end": "16:40",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Davidia 1"
  ],
  "title": "How to get people on board with responsible tech ? From awareness to engagement",
  "speakers": [
   {
    "name": "Jérôme Vigogne",
    "title": "Consultant",
    "company": "ACT4CHANGE"
   },
   {
    "name": "Sandra Sydow",
    "title": "President & Cofounder",
    "company": "Planet On Stage (NGO)"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Engagement in responsible tech practices; sustainability.",
  "abstract": "We'll share with you key insights on how to improve people's engagement on adopting responsible tech practices. At Planet On Stage, with more than 6 years of experience in running awareness programmes in companies, universities and communities worldwide, we have managed to engage 25k+ people on the difficult topics related to environmental challenges. How did we achieve this? Combining powerful visual storytelling and collective experiences through our unique Digital Impact Pitch format, with Communication with Impact Masterclasses that equip people with the soft skills to inspire change. Join us to discover how to make responsible tech everyone’s business!",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4587,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "16:20",
  "end": "16:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 6"
  ],
  "title": "Fair & Responsible AI for Sustainability: Governance, Resilience and Adoption",
  "speakers": [
   {
    "name": "Louis de Diesbach",
    "title": "Technology Ethicist",
    "company": "Euroclear"
   }
  ],
  "facets": [
   "accountability",
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Responsible AI governance and footprint; not API surface.",
  "abstract": "Every organisation is rolling out AI, but few can answer: AI, sure - but to do what? Too often adopted as an end rather than a means, AI brings runaway costs, governance blind spots, and a growing environmental footprint. This talk offers a practitioner's roadmap for responsible AI adoption: an AI vision anchored in your mission, a governance ecosystem across four pillars (strategy, risk, support, continuous improvement), and three practical starting points - trustworthy data, a first AI policy, training people. We'll close with the questions that matter: what do we do with the productivity",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4553,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "16:20",
  "end": "16:45",
  "track": [
   "GreenIO"
  ],
  "room": [
   "Dragonnier 5"
  ],
  "title": "Une IA sobre qui doit tout prouver",
  "speakers": [
   {
    "name": "Jérémy Pastouret",
    "title": "Co-founder and web developer",
    "company": "Les e-novateurs"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "Frugal AI reading privacy policies; methodology, not API surface.",
  "abstract": "Personne ne lit les politiques de confidentialité. Nous avons voulu en faire lire 134 à une IA, et les questions sont arrivées avant les réponses. Tout confier au modèle, ou seulement une partie ? Est-ce qu'un simple algorithme, ou des statistiques, font aussi bien sans IA ? Comment savoir quand le modèle invente ? Combien d'appels sont vraiment nécessaires ? Que reste-t-il aux humains ? RETEX sur les choix faits, nos erreurs, nos succès et ce qu'un usage sobre de l'IA veut dire concrètement.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4418,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "16:25",
  "end": "16:50",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 2"
  ],
  "title": "Integrations should be regenerated, not maintained",
  "speakers": [
   {
    "name": "Predrag Radenkovic",
    "title": "CTO",
    "company": "Codeplain"
   }
  ],
  "facets": [
   "contract_quality",
   "developer_ergonomics",
   "agent_readiness"
  ],
  "agent_readiness_dimensions": [
   "spec_presence"
  ],
  "why": "Regenerating integrations with coding agents rather than maintaining them.",
  "abstract": "Coding agents have changed the economics of software development. Developers should expect their organizations to own an order of magnitude more code than today. But unless agents can also maintain software reliably, with humans only supervising the process, the gains from AI will be limited. Building on Chad Fowler’s idea of Regenerative Software, this talk explores a novel approach: if a coding agent can repeatedly regenerate software from durable information capturing what must not change, that shows the agent understands what it is building. That understanding enables the agent to maintain the software reliably as it grows and evolves, instead of having the agent’s performance degrade as misunderstandings accumulate in the code through incremental maintenance. We'll explore the topic on practical example of API integration development and maintenance, and how this approach already works in practice.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4843,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "16:25",
  "end": "16:50",
  "track": [
   "FINOS"
  ],
  "room": [
   "Acacia 6"
  ],
  "title": "Networking reception",
  "speakers": [],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "Networking reception, no API relevance.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4860,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "16:25",
  "end": "16:50",
  "track": [
   "Apidays"
  ],
  "room": [
   "Aubépine 1"
  ],
  "title": "De 6 mois à 2 jours : la révolution LLM pour le traitement documentaire",
  "speakers": [
   {
    "name": "Vinorth Varatharasan",
    "title": "Senior Data Scientist",
    "company": "Kolecto"
   }
  ],
  "facets": [
   "outside"
  ],
  "agent_readiness_dimensions": [],
  "why": "LLM document processing; no API surface.",
  "abstract": "Un nouveau document arrive, un fournisseur change son template, une réglementation impose un nouveau champ : et c’est reparti pour un tour de réannotation et de réentraînement. Pour une équipe qui gère des dizaines de types de documents, ce cycle ne s’arrête jamais, mobilisant plusieurs ingénieurs ML pendant des mois pour un gain marginal. Ce cycle vient du deep learning classique, qui doit réapprendre chaque mise en page avant de pouvoir la lire. Les LLM multimodaux savent déjà lire un document. Le travail change de nature : on n’entraîne plus de modèles sur de gros GPU, mais on configure un LLM avec un prompt, un schéma de sortie et des règles métier. À partir d’un cas réel de vérification d’identité bancaire (KYC) et d’une démo live, nous verrons comment passer d’un POC en deux jours à un système en production. Le prompt devient le nouveau modèle, mais chaque sortie doit être validée avant qu’on lui fasse confiance, et un humain reste en charge des cas douteux.",
  "url": "https://www.apidays.global/events/paris#agenda"
 },
 {
  "id": 4678,
  "day": "2026-12-03",
  "type": "Talk",
  "start": "16:40",
  "end": "16:50",
  "track": [
   "Karate Labs 2nd User Conference"
  ],
  "room": [
   "Aubépine 7"
  ],
  "title": "Close: conclusions from the day",
  "speakers": [
   {
    "name": "Kapil Bakshi",
    "title": "CEO/Co-founder",
    "company": "Karate Labs"
   },
   {
    "name": "Pier-Jean Malindrino",
    "title": "CTO",
    "company": "Scub"
   }
  ],
  "facets": [],
  "agent_readiness_dimensions": [],
  "why": "No abstract; closing remarks.",
  "abstract": "",
  "url": "https://www.apidays.global/events/paris#agenda"
 }
]