Running Azure SQL at cloud scale means operating one of the largest and most complex database platforms in the world-under constant customer load, strict reliability guarantees, and intense time pressure during incidents.

In this session, we will share how the SQL engineering team at Microsoft is using AI to fundamentally change how SQL runs SQL.

We will walk through how Bluebird, Microsoft’s deep code-aware AI platform, and SQL Live Site Agents are being used together to help engineers detect, investigate, and mitigate live site incidents faster and with greater confidence. These agents bring together source code understanding, production telemetry, incident context, and operational knowledge into a single, guided workflow-directly in the tools engineers already use.

You’ll see how we are moving beyond traditional runbooks and fragmented tools toward AI-assisted workflows that reduce cognitive load, shorten time-to-mitigation, and make expert-level troubleshooting accessible during high-severity incidents.

We’ll also cover how we design guardrails such as human-in-the-loop approvals for high-risk actions, ensuring trust, safety, and auditability in mission?critical systems.

This session is grounded in real production learnings from operating Azure SQL and is relevant to anyone running large-scale, always-on data platforms.

Attendees will leave with practical insights on building trustworthy AI agents for operations, and how similar patterns can be applied in their own environments.