Modern resiliency from build to recovery through Agentic AI | BRK228
Rochak Mittal, Shobhit Garg, and Adity Agarwal present a Build 2026 breakout on treating resiliency as an agent-first practice, showing how an agentic AI assistant can connect build, operations, troubleshooting, and recovery workflows across repos, dashboards, runbooks, and collaboration tools for Azure workloads.
Overview
The session argues that resiliency failures are often caused by fragmented tooling and context switching during incidents (for example, bouncing between source repos, dashboards, runbooks, and chat/collaboration tools). It reframes resiliency as an end-to-end practice that spans:
- Build and validation
- Operate and monitor
- Troubleshoot incidents
- Recovery actions
The presenters demonstrate an agentic AI approach aligned with Azure’s agentic model, where an assistant links workflows across IDEs, source repositories, and collaboration tools to speed up validation, operations, and recovery for a critical workload.
Session chapters (from the video description)
- 0:00 - Introduction and overview of resiliency discussion
- 00:10:00 - Demo: Building a resilient application using Azure Copilot and a resiliency agent
- 00:13:02 - VMSS top recommendations for zonal resiliency and alternate options
- 00:13:53 - Cross-cutting considerations and summary checklist for resilient applications
- 00:16:26 - Demo: Generating resilient infrastructure from scratch using Azure Copilot
- 00:22:09 - Conclusion of first phase: Start resilient
- 00:33:55 - Tracking completion progress and downloading custom action plans
- 00:35:20 - Demo setup: Java Ask HR app and growing dependencies
- 00:43:14 - Introduction of Azure Chaos Studio Workspaces and public preview announcement