Product Pieces: AI Solutions: The plain-English field guide to building with — and how every idea goes wrong.

Everyone wants to "add AI." Almost nobody can say what happens when it's wrong. This is the plain-English field guide for the people who have to answer that - founders, product managers and leaders building with AI for real, often with money, people or a regulated process on the line. Here's the fact the whole book turns on: a normal feature does exactly what it's told, every time. An AI feature produces a best guess that's usually right and sometimes confidently wrong. Demos are cherry-picked; production isn't. Almost everything that goes wrong with AI products - the frightening bill, the made-up answer, the model that aced the demo and fell over in the wild - follows from that one shift. Product Pieces: AI Solutions takes the jigsaw approach of the first book and rebuilds it for AI. You start with the corner pieces - what AI actually is, deterministic versus probabilistic, why models hallucinate - then work outward to the harder, higher-stakes ones: RAG versus fine-tuning, evals as the new unit test, observability and drift, cost per successful task, agents, audit trails, explainability, and governing AI when the stakes are real. Every one of the 113 concepts is a single page, the same seven beats each time: the idea in one line, what it is in plain language, when to use it, where you've already seen it, how it goes wrong, what it connects to, and one small thing to try today. No maths to wade through. Analogy first, jargon second. The aim isn't to turn you into a machine-learning engineer. It's to make you useful and unbamboozled fast - able to ask the right questions, spot the trap before you fall in it, and know when not to use AI at all. The companion to Product Pieces. Read either first.

Pagina's: 212, Paperback, Independently published