ATDD with AI Agents

AI coding agents can write a lot of code, fast. But that code might not behave the way you need — or it quietly breaks something that used to work. The hard part isn't getting code; it's knowing that it works correctly and hasn't introduced a regression.

Acceptance Test-Driven Development (ATDD) answers that. You agree on concrete examples of the behavior you want before any code is written, turn them into executable acceptance tests, and let those tests define "done." This book shows how AI agents change — and amplify — that practice: the tests become the specification you hand the agent, and the safety net that proves it delivered.

You'll move from traditional ATDD to a modern ATDD practice built around AI agents — from the fundamentals, through writing acceptance tests precise enough to drive an agent, to a full worked case study.

By the end, you'll be able to apply the full ATDD-with-agents workflow to implement a user story from start to end:

• Use AI agents to write acceptance tests based on acceptance criteria — and approve the resulting specification yourself
• Use AI agents to implement code that satisfies the acceptance tests — and review the code yourself
• Keep agents under control — constrain their scope and optimize their token efficiency

Who this book is for

Software developers, tech leads, and QA engineers who already use — or want to use — AI coding agents, and want a disciplined way to get reliable results from them. This book assumes a basic understanding of ATDD and a basic familiarity with AI agents; you don't need to be an expert in either.

A note on this book — I'm writing this book with an agile approach — the same iterative way good software gets built. Instead of waiting months or years for a finished book, you get instant access today, a say in where it goes through your feedback, and every update for free as it ships. It's a living book, so things will change along the way — the outline, the chapters, the schedule — and I may add, drop, or rework topics as I learn what helps readers most. So please pick it up for what it is today, not as a promise of exactly what it'll become.