What this book is#
This is a practical account of how I actually work with AI agents as a data engineer, written so I can hold myself to it and so anyone doing similar work can borrow what fits. It is not a survey of every tool, and it is not a promise that the machine will do the thinking for you. It is a record of the habits, the setup, and the guardrails that let me hand real work to an agent and still stand behind what ships.
The through-line is that AI is an accelerator, not a substitute for judgment. It pays off when you already understand the problem well enough to verify the answer, and it becomes a liability the moment you let it run somewhere you cannot follow.
There is a personal reason I am writing it now. After more than a decade of building data systems, I want to take the next step in my own craft, the one that turns hard-won practice into something I can articulate clearly enough to be useful to someone else. Writing this down is how I make that move, from doing the work well to being able to teach it, and that ambition is as much a part of this book as any technique in it.
Who I am writing for#
I am writing first for the engineer I was a few years ago, capable at the craft but new to letting a model touch the work. If that is you, this book is meant as company from someone a little further down the same road, not a lecture from someone who never made the mistakes.
It assumes you can already read a query, reason about a data model, and recognize when an answer is wrong. The faster you can verify, the more this way of working gives back.
How I will write it#
I write in the first person because the choices and opinions presented here are most certainly mine. When the book turns toward setup or habit, I will often speak in the second person, because the reader I have in mind most often is still a version of me.
I will use concrete examples from real work, including the times an agent took a liberty I did not ask for and the recovery that followed. Tools and model names will change underneath this book, so I will keep the durable principles in the foreground and treat the specifics as illustrations, not the point.