I sit down with Till Döhmen, AI Lead at MotherDuck, to explore the evolving landscape of AI-assisted SQL generation. Till brings a unique perspective from his PhD research in databases and his work building AI features at MotherDuck, the serverless data warehouse built on DuckDB.
We discuss how text-to-SQL has matured from academic benchmarks to practical tools, the importance of documentation and schema context for agents, and the emerging role of MCP servers and skills in customizing agent behavior. Till shares insights on building trust in agent-generated queries, why thinking about what a new hire would need to know helps you build better AI workflows, and how agents are reshaping the analyst's role from SQL writing to question refinement.
