If your APIs aren’t agent‑ready, your AI agent is just chat: [...]
summary
Central claim: agent autonomy collapses if APIs are brittle — reliability compounds across tool calls so 95% success over ~20 calls yields ~36% end-to-end reliability. Concrete deployments: a data governance agent running in the Microsoft stack for a major German energy provider, plus QA, monitoring, insights and coding agents closing the improvement loop. Practical fixes: full OpenAPI schemas, machine‑readable errors, idempotency keys, explicit side‑effects, enforcement via an AI gateway, and distributed tracing with OpenTelemetry for replay, attribution and alerts.
AI-written summary · about
chapters (9)
- 0:00Agent-ready APIs intro
- 1:10Agent autonomy levels
- 5:44Closing loop with improvement
- 10:46API reliability defines autonomy
- 13:57Bad APIs hinder agents
- 17:04Six ways to design agent-friendly APIs
- 18:28Guardrails for agent safety
- 21:43Traceability as third pillar
- 23:52Agent readiness check
related talks
32mYour APIs Are Not Agent-Ready: How to Test What AI Agents Actually Do
35mUnleash Your Inner API Architect: Build Agents That Deliver
13mAre Your APIs Ready for AI Agents?
22mBusiness Wants In—Here’s How to Drive API Wins (and Keep Your Sanity)
24mAPIs Are the Nervous System of Agentic AI - Designing Agent-Ready APIs for Aut…
24mThe API Economy's Next Chapter: AI Agents as First-Class Clients
29mAgent-Ready Orchestration: Engineering the Execution Layer.
28mAI and the state of APIs in 2026