05:35Take AI agents from prototype to production with OpenTelemetry | ODSP909
From the creator
Shipping an AI agent is easy. Trusting it in production is hard. In this session, build a multi-agent travel planner using the Microsoft Agent Framework, then instrument it end-to-end with OpenTelemetry and New Relic. You'll learn how to trace agent decisions, monitor response quality, catch prompt injection attacks, and build CI/CD quality gates—so your AI systems are observable, secure, and ready for real customers. 𝗦𝗽𝗲𝗮𝗸𝗲𝗿𝘀: * Harry Kimpel 𝗦𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻: This is one of many sessions from the Microsoft Build 2026 event. View even more sessions on-demand and learn about Microsoft Build at https://build.microsoft.com ODSP909 | English (US) | Agents & apps Pre-recorded | (100) Foundational #MSBuild Chapters: 0:00 - Question: Why do AI agents behave unpredictably in production? 00:01:07 - Concrete example: Travel planner output error (Kyoto vs Tokyo) 00:05:06 - Challenge: Limited Visibility into Agent Decisions 00:05:49 - Built-in OpenTelemetry Support in Microsoft Agent Framework 00:06:56 - Viewing Agent Trace Data in New Relic 00:09:30 - Demonstrating log-to-trace correlation in New Relic 00:10:16 - Implementing quality gates for AI outputs 00:11:52 - Introduction of Microsoft Foundry guardrails at platform level 00:13:31 - Recap: Making AI agent behavior observable through telemetry and guardrails
Published 2026-06-03 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.





