Skip to content

BRK415 Breakout Wed 18 Nov 11:30 PST

Put AI agents to work in Azure cloud operations

Brendan Burns, Ruhiyyih Mahalati

Summary

From Microsoft's AI summary

Brendan Burns and Ruhiyyih Mahalati presented "agentic operations": agents working within guardrails across the software lifecycle. The ARM MCP server is in public preview. SRE Agent has been generally available since March; new preview features include Teams and Slack integration, multi-cloud incidents and customer-managed keys. Azure Policy rules now apply to agents, demonstrated blocking Key Vault creation in production. The Azure Copilot deployment agent and Azure Monitor health models are newly GA. New-generation Azure CLI and PowerShell are in public preview; SAP and Oracle workload agents are in limited preview. Caveat: based on Microsoft's AI summary, not the video; no pricing or regional availability stated.

Session description

Microsoft's catalogue description

Move from AI assistance to agentic cloud operations. Through live demos, see Azure agents use rich operational context to investigate incidents, uncover cost optimization opportunities, and take corrective action with granular controls. Explore new Azure Copilot experiences and Azure Resource Manager MCP Server enhancements that give agents deeper environmental awareness. This session is built for FinOps, SRE, platform engineering, and AI infrastructure teams.

Demo timestamps

Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.

  1. Skip to 0:10:50

    ARM MCP server in Visual Studio Code

    VS Code with the ARM MCP server lists resources in a fictional Caldova subscription. The server generates, validates and runs Azure Resource Graph queries from natural language and drills into resource properties.

  2. Skip to 0:15:57

    Azure Copilot analyzing a cost spike and budget alert

    In a FinOps role, a budget-threshold alert is opened and Azure Copilot analyzes the spike. Adding application business context makes a repeated question return more specific cost insight.

  3. Skip to 0:25:55

    SRE agent autonomously mitigates a GitHub-filed incident

    The SRE agent opens a GitHub issue with root-cause analysis after order processing becomes unavailable. A stopped back-end database server is restarted once the operation is approved, with mitigation in about seven to eight minutes.

  4. Skip to 0:27:41

    SRE agent tool guardrails

    Settings show which tools the SRE agent may run, which need operator approval, and which are denied, such as a production delete.

  5. Skip to 0:28:28

    SRE agent review scenario and deployment guardrail workflow

    A second alert is handled with the operator driving mitigation. A later deployment-guardrail workflow, run by HTTP trigger or schedule, reports synthetic tests with 160 of 160 HTTP 200 responses and latency checks.

Watch on-demand (opens in a new tab)

← All sessions