Unattended An AI Operator's Notebook

This book wasn’t planned. It started as notes.

In January 2026, I gave an AI agent access to my email, my calendar, and my messaging platforms. Not a chatbot, an autonomous system that could read, write, and act on my behalf without me sitting at the keyboard. I wanted to see what would happen when I stopped treating AI as a tool I picked up and put down and started treating it as something that worked alongside me.

What happened was seven months of the steepest learning curve I’ve had since I got out of the Army and walked into my first IT job. Except this time, nobody had written the manual. There were no best practices I could find, no vendor certifications, no “AI Operator” job title on LinkedIn. Just other people on the bleeding edge, trading notes on social media about what did and didn’t work, a handful of language models, and a growing list of things that went wrong in ways none of us expected.

I wrote it all down. Not because I thought it would become a book, but because I needed to remember what I’d learned before I forgot it. The notes turned into LinkedIn posts. The posts turned into a book.

Why this book exists

I’ve been in technology for over thirty years. Web development and the beginning of the internet in the nineties. Enterprise automation. Infrastructure. Security. I’ve watched entire job categories appear, evolve, and disappear. I’m watching it happen again right now.

The tools coming our way aren’t incremental improvements; they are a different way of working. And the people who figure out how to use them, really use them, not just ask a chat LLM to write an email, are going to have an enormous advantage over the people who wait for someone to hand them a training manual.

That manual doesn’t exist yet. This book is my attempt to start writing it.

What I’m describing in this book isn’t just AI agents. It is something more specific: an AI Orchestration Agent, or AOA. That’s the term I use here for the coordinating agent, the one that runs the others. This book is about how operators, not engineers, can put one to work. A chatbot answers questions. An agent performs tasks. An AOA coordinates across tools, platforms, and other agents, reading email, checking calendars, posting to social media, managing deployments, and handing work off to specialists, all without a human directing every step. That distinction matters, and it is the foundation of everything in this book.

I didn’t write this for AI researchers or machine learning engineers. I wrote it for the person who runs things. The IT director who keeps the lights on. The small business owner who handles everything because there’s no one else. The operations manager who knows what “works in production” actually means.

I also wrote it for the student or recent grad staring at a job market that doesn’t look like the one I walked into. You can understand this method without being a programmer. Running it unattended, in production, still takes technical judgment or someone who has it; I’ll tell you where.

If this book helps someone find a rung on a ladder that keeps changing shape, it was worth writing.

How it got here

This book was written the way I think it should be read: one lesson at a time, grouped by the operating problem rather than strictly by date. Chapters 1 through 3 are the story: how the project started, what I brought to it, and the first real test. Chapters 4 through 10 are the lessons: what broke, what worked, and the principles I extracted from both. Chapter 11 is the blueprint, Chapter 12 is what it takes to run it, and the Afterword is how this book was made. The appendices are reference material: a glossary and a set of starter scripts to get you moving.

Every chapter is based on something that actually happened. I kept the daily logs, the terminal output, the Discord messages, the memory files. Where I made mistakes, and I made plenty, I left them in. You learn more from the failures than from the victories, and I’d rather be useful than impressive.

The agents that run my operation today, Tom, Oscar, Cash, and a handful of others, were all built during the period this book covers. By the end of the book, they are a team. But they started as experiments, and most of the interesting lessons came from the experiments that didn’t go the way I expected.