Leveraging Azure AI Foundry to Build Intelligent Release Agents
summary
Builds an end-to-end release agent using Azure AI Foundry (chooses an AI hub resource for production), connects to Application Insights via its API with a bearer token to fetch logs, runs an LLM-based analyze-logs prompt that returns a boolean anomaly flag, and uses the Azure DevOps API to trigger pipeline and loop-check status until success or failure. Deployed flows produce an API, run serverless tests, monitor for 15 minutes post-deploy, and the demo reports no anomalies; target scale is ~15 data centers, 4 pipelines per region, ~750–1,000 services.
AI-written summary · about
chapters (13)
- 0:00Introduction and speaker background
- 1:02Problem with current DevOps setup
- 2:05Introducing Azure AI Foundry
- 3:28Creating Azure AI Foundry projects
- 5:36Understanding prompt flows
- 6:20Release agent architecture overview
- 9:13Testing flows in Azure AI Foundry
- 10:39Deploying to production
- 11:22Chat flow and pipeline triggering
- 13:17Live demo of intelligent release agent
- 14:21Microsoft Clarity case study
- 15:47Benefits and future improvements
- 16:08Q&A session
related talks
22mMaking AI Accountable: Azure AI Foundry
30mAI-Augmented DevOps with Platform Engineering
05h 17mBuilding AI Applications
01h 54mBuilding Code First AI Agents with Azure AI Agent Service
01h 00mGetting started with Azure DevOps
39mSimplify, speed and improve development with DevOps
36mBuilding AI Agents in .NET with Azure AI Foundry
07h 25mAI Agents Conference 2026 - Day 2