This book is licensed for the personal use of the buyer. Please do not share or resell the file. Short quotations with attribution are welcome.
The information in this book is provided for education. It is not legal, financial or other professional advice. Product and company names are used only to describe them and belong to their owners.
This is a free sample: the introduction and first chapter.
IntroductionIntroduction
AI products behave differently from ordinary software. A button either works or it does not; a model is right most of the time and wrong in ways that are hard to predict. Ordinary features are built once and stay the same; AI features depend on data that changes, and their quality must be measured continuously. Ordinary specifications describe exact behaviour; AI specifications describe acceptable distributions of behaviour.
Product managers need new habits to build AI products people trust and use. This book walks through the whole life cycle: choosing problems where AI genuinely adds value, defining quality bars, working with data and model teams, designing for uncertainty and errors, evaluating before and after launch, managing costs and pricing, and handling risk and compliance. Case studies show common traps and how strong teams avoid them.
No coding is required, but you will learn to ask the right technical questions.
How the book is organised
Chapters 1 to 10: what makes AI products different, and choosing the right problems.
Chapters 11 to 21: defining quality, data, working with ML teams and build-versus-buy decisions.
Chapters 22 to 33: designing for uncertainty, trust and feedback.
Chapters 34 to 46: evaluation, launch, metrics, costs and pricing.
Chapters 47 to 56: risk, regulation, fairness, safety, transparency, reviews, teams, rituals, requirements and stakeholder communication.
Chapters 57 to 61: a worked quality bar, measuring time saved, discovery interviews, competitive strategy and retiring features.
Chapters 62 to 68: case studies and checklists.
Chapter 1What Makes AI Products Different
Ordinary software
AI products
Deterministic behaviour
Probabilistic behaviour
Specified by rules
Learned from data
Tested once, stable
Measured continuously, drifts
Errors are bugs to fix
Errors are expected; manage their rate and impact
Cost mostly fixed
Cost may scale with usage (inference)
You have reached the end of the free sample. The full book continues with 67 more chapters, plus the appendices.