In late 2022, the "Chat GPT moment" made foundation models widely known, but a less discussed use case emerged: analytical AI, where AI decides rather than creates. This handbook focuses on using AI for measurable, specific tasks, contrasting with generative uses. It emphasizes building ground-truth datasets, running on smaller models for efficiency, and accepting higher latency, similar to OLAP vs. OLTP data processing.

Written for data, ML, and analytics teams, it covers primitives, patterns, and architectures for production deployment, aiming to be a living resource for developers.