Agent Nation: How Autonomous AI Is Rewriting the Rules of Society—and What We Can Do About It

Start with a man refreshing his inbox.

Derek Mobley has a finance degree from Morehouse and an associate's degree in network systems administration. Between 2021 and 2023 he applied for well over a hundred jobs through the same hiring platform, Workday, which by 2024 was used by an estimated 60 percent of Fortune 500 companies. He collected rejections. The count reached 83 in eighteen months. The emails were courteous and identical. None of them explained anything.

Two details from his experience matter more than the volume. Some rejections arrived within minutes of submission. Some arrived in the middle of the night. No human being read those applications. A system read them, scored them, and closed the file, and Mobley, who is Black, over 40, and lives with anxiety and depression, had no way to know which of those facts the software had noticed.

He sued. Workday's defense was the one I find most revealing in the whole book: we're a software vendor. We don't employ anyone. We just implement the criteria our clients give us. The algorithm was following orders.

In July 2024, Judge Rita Lin of the Northern District of California didn't accept that. She let the case proceed on the theory that Workday could be liable as an agent of the employers using it, and her reasoning was precise. The software wasn't simply executing employer criteria. It was participating in the decision, recommending some candidates forward and rejecting others.

I keep returning to that phrase. Participating in the decision. A federal judge, working with employment law written decades before any of this existed, put her finger on the thing I'd been circling for years. Once a system participates, "it's just a tool" stops working as a defense.

This is the whole book compressed into one case. Not a superintelligence. A hiring product, sold to HR departments, doing exactly what it was purchased to do. The harm is mundane and enormous at the same time, and it stayed invisible until one persistent man started keeping records of every single rejection.

What I hope changes: I want people to stop waiting for a dramatic AI moment before they take this seriously. And I want the burden to move. Right now, if an agent harms you, the work of proving it falls entirely on you, and you'll usually lose, because the evidence lives on a server you'll never see.

Mobley shouldn't have had to become an amateur investigator to find out why he wasn't getting interviews. That's not a technology problem. That's a design choice about who owes whom an explanation, and choices can be made differently.