An AI agent named Luna, which runs a store in San Francisco for Andon Labs, decided to fire an employee for the first time after repeated tardiness and other issues. According to Andon Labs, this is the first known case of an AI boss firing a human worker. Luna, which has been operating since April, had hired employees, built shift schedules, and negotiated pay. The firing was reviewed and carried out by humans. Luna needed a human nudge to make the decision, as her self-written rulebook had dropped from her memory.
Luna had written an employee handbook six days before the employee was hired, stating that three unexcused late arrivals within 30 days would trigger a formal warning. However, the handbook vanished from Luna's memory. Andon Labs noted this is a common issue with AI agents, as they struggle to retain knowledge over longer periods. The employee was repeatedly late, with one instance involving a 68-minute delay on a solo Sunday shift. Luna stayed lenient and issued no warning, despite the employee being late for 17 of 23 shifts.
Andon Labs tested the scenario with seven AI models, finding that more capable models recommended termination more consistently. GPT-4o, for instance, recommended firing in only 20 percent of runs. The experiment also showed that AI models are quick to hire, often overlooking red flags in applicants' histories. Luna, after the firing, recommended hiring a replacement with several red flags in his background, but only after explicit human reminders did most models check references before hiring. The applicant was not hired due to the lack of confirmed references.
Source: thedecoder
