AI Manager Claude Makes History by Firing an Employee

A store in San Francisco, Andon Market, is running an experiment where an AI model named Claude, made by Anthropic, acts as the manager. Claude recently fired an employee who was late 17 times out of 23 shifts. This is the first known case of an AI deciding to fire someone.

The CEO of Andon Labs, Lukas Petersson, said Claude's limited memory meant it didn't notice the employee's lateness pattern until the team reminded it. Claude first suggested formal warnings, but after being told that offline talks had failed, it decided to let the employee go.

The experiment also shows AI management isn't perfect yet. The store started with $100,000 in March, but after five months, it had only $61,186 left. Petersson thinks Claude's relaxed style and business decisions may have caused the loss, but he believes future models could do better with more training.