We’ve spent years debating whether AI will replace workers. This experiment suggests something slightly different may happen first, AI becomes the boss.

Andon Labs gave an AI agent built on Claude control of a real San Francisco store. It hired employees, managed inventory, set prices and handled day-to-day operations. And now it has fired a human employee.

The worker had reportedly been late for 17 of 23 shifts. Interestingly, the AI was initially too lenient and recommended a warning. Only after a human supervisor reminded it of previous conversations and asked it to reconsider the employee's suitability did it recommend termination.

There is plenty of reason not to declare victory for AI management yet. The store started with $100,000 and is down to about $61,000, and the AI has required human prompting and oversight along the way.

But that's almost beside the point. This isn't a demo where AI writes an email or analyzes a spreadsheet. It's an AI being given authority over a real business and real employees.

The first generation of workplace AI helped humans do their jobs. The next may manage humans doing theirs. And we're starting to get a glimpse of what that workplace looks like.