Agentic AI
Azure SRE Agent automates incident response while keeping humans in control
Microsoft's new agentic operations platform helps site reliability engineers cut toil by automating root cause analysis, remediation, and proactive monitoring across infrastructure and services.
The New Stack
Picture this: an alert fires at 3 a.m., but by the time an engineer wakes to the page, an AI agent has already analyzed telemetry, pinpointed the root cause, and prepared a fix waiting for approval. Azure SRE Agent aims to transform how teams respond to production incidents by automating the repetitive investigation work that consumes hours of engineering time.
Inside Microsoft, over 3,000 service teams now lean on the platform to handle incident triage, root cause analysis, and mitigation—managing more than 1.8 million incidents, many resolved in minutes. At customer InEight, the agent identified a performance bottleneck in one of the company's 14 products and correctly recommended scaling Redis rather than the app service, a diagnosis that manual inspection had missed.
How agents are already operating in production#
The capabilities have reached a point where some Microsoft internal teams rely on autonomous agent management for more than half their incidents. According to Shamir Abdul Aziz, lead program manager for Azure SRE Agent, these are "safe" operations—service restarts, scaling decisions, rollbacks, and change escalations—where humans set governance rules upfront and the agent executes within those constraints.
Mehta highlights one concrete example: when a PyPI package upstream broke an internal dependency, the agent immediately spotted the cause as it deployed to the first region, then suggested a rollback and identified missing tests in seconds. Such proactive detection requires reasoning beyond deterministic queries; agents can assess whether a large production payload carries degradation risk before issues cascade.
Context and harness engineering#
Raw AI power alone isn't enough. Azure SRE Agent combines what the team calls "harness engineering"—grounding the agent in your infrastructure, code, audit trails, and institutional knowledge—with production-grade governance, verification, and telemetry.
The platform provides native integrations with Azure Monitor, Application Insights, Log Analytics, and Azure Resource Graph. Beyond Azure, managed connectors link Azure DevOps and GitHub, while MCP connectors bring in Google Drive, Confluence, and third-party systems. Teams can codify runbooks and skills as Markdown artifacts in version control, making them reviewable, testable, and improvable over time—even by other agents.
