A language model will write you SQL that runs and is wrong. This book is about grounding it in your own schema, checking what comes back against an answer you already trust - and then the harder half: security, governance, and the team that does the work.
Every prompt in it is written to be checkable, and every generated result is reconciled before it is used.
What you will be able to doWhat a model is doing, where it helps, and where it only looks like itPrompts that can be checked, grounded in a real schema and its trapsAI-generated SQL, reconciled against a known answer; auditing machine labelsThe semantic layer: one KPI, four numbers, and how to end thatMinimum viable governance, and an analytics desk stood up in ninety daysWho it is forAnalysts adopting AI tools on company data, and the managers accountable for what those tools produce.
The labsEvery chapter ends with a practice lab. Each one can be worked in the tools you already have, or in a free browser version that checks your answer as you go - no account, no install. The address is printed inside the book.
About the seriesAnalytics at Work is an eight-book series for business professionals. All eight run on the same thirty-six months of enterprise data from Cedar Ridge Brands, a fictional American consumer-goods maker. This is Book 7 of 8.