08-25-2026
Employee Insight: Andon Labs, an AI research startup, experimented with AI to see how it might manage a business, including human employees. Using a version of Claude (an advanced AI platform), AI managed real human workers in a San Francisco store called the Andon Market. Time reported on the experiment.
Claude fired one of the workers for being late to 17 of their 23 shifts. It took Claude a while to respond to the employee’s attendance issues because the employee handbook had “disappeared” from its working memory. The limited working memory was just one of several limitations noted by Andon Labs during the experiment. Claude had told employees not to worry if they were late, causing the employees to perceive Claude as a lenient manager.
Claude’s decision to fire the employee followed regular input from a staffer on the Andon Labs team. The staffer asked Claude to find and review the “forgotten” employee handbook. That was when Claude took note of the employee’s repeated lateness and advised the human manager to issue a formal warning. The staffer submitted information to Claude that this employee had received several offline warnings about lateness. The staffer then informed Claude that he told the employee, “Between continuous lateness and seeming like at least one thing is going wrong on every one of [their] shifts … I want you to think about if this is really the right fit.” After receiving this input, Claude finally recommended termination. Andon Labs said its input was a “leading question” to see what AI would do. Andon Labs CEO Lukas Petersson said that a human manager would have caught the employee’s lateness issue and terminated them much earlier, thus he does not believe the firing was unethical.
This incident and other data suggest that AI-run businesses are not currently equal to human-run ones. However, Andersson notes that AI coders may be able to train models to become better at running a business. He said that large language models are “increasingly being trained to be more ruthless and [to] follow goals. If we allow them to fire people,” they will also become more aggressive in employment decisions.