The AI-assisted DBA: performance triage with GitHub Copilot and the MSSQL extension
Can an AI assistant read an execution plan? Sometimes. This session puts GitHub Copilot, the MSSQL extension for VS Code and the SQL MCP server to work on real performance problems: a parameter-sniffing regression, an implicit conversion hiding in a join, a missing-index recommendation that would make things worse, and a blocking chain. For each case we see what Copilot diagnoses correctly, where it confidently gets it wrong, and the context techniques that shift the balance, such as supplying plan XML, wait stats and schema rather than asking cold. The result is a repeatable triage workflow where the assistant does the first pass and the DBA keeps the judgement. You leave knowing which parts of the job to hand over, which to keep, and how to tell the difference before it costs you a production incident. What attendees will learn 1.A four-step AI-assisted triage workflow for SQL Server performance issues 2.The specific failure modes of LLM plan analysis and how to catch them 3.How to wire Copilot to live database context with the MSSQL extension and MCP
Speaker
Gethyn Ellis
Where data strategy meets real-world execution.
Gethyn is a Microsoft Data Platform MVP and consultant helping organisations modernise their data platforms with SQL Server, Azure, Microsoft Fabric, Power BI, and AI. He leads a UK consultancy that supports clients across Europe and North America with migrations, high-availability design, performance engineering, analytics strategy, and data governance. A Microsoft Certified Trainer with over 200 courses delivered, he speaks widely on data transformation, analytics, and the future of AI-powered... read more