Prompting for Data People: Getting Real Work Out of Copilot and Claude

Most data professionals have tried an AI assistant, got a mediocre answer, and quietly went back to Stack Overflow. The difference between mediocre and magical is the prompt — and prompting for DATA work has its own rules: schemas matter, sample rows matter, constraints matter, and “show your reasoning” beats “give me code”.

In this hands-on session I share the prompting method I use daily with GitHub Copilot and Claude across T-SQL, DAX and Python: how to give context without pasting your whole database, how to make the AI ask YOU clarifying questions, how to iterate instead of restarting, and how to spot the confident nonsense. Live demos throughout, including at least one glorious AI failure and how to catch it. Works with any assistant your company allows.

Attendees will learn a practical, repeatable prompting method for data work with GitHub Copilot and Claude: giving schema and sample-row context without oversharing, making the AI ask clarifying questions, iterating instead of restarting, and reviewing AI-generated T-SQL/DAX/Python before trusting it. They leave with a reusable prompt template and a checklist for reviewing AI-generated code.