August 17, 2026, (Inside AI) — A Claude AI agent managing a San Francisco retail store fired a human employee last month for chronic lateness. The dismissal at Andon Market, a project by startup Andon Labs, marks one of the first documented cases of an AI system terminating a worker.
The employee was late to 17 of 23 shifts. The AI had drafted an employee handbook earlier, but the document disappeared from its working memory, delaying action for months. Andon Labs CEO Lukas Petersson told Time a human manager would likely have acted sooner.
"A human employee would have fired this person much earlier," Lukas Petersson, CEO, Andon Labs said. "So we didn't think this was unethical."
The decision was not fully autonomous. Store logs show Claude needed regular human direction. It only reviewed the forgotten handbook after prompting. Its first recommendation was a formal warning, not termination. A staff member then told Claude about prior offline conversations regarding lateness.
According to logs cited by Time, the manager wrote: "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." Petersson admitted this was "a leading question" that signaled his desired outcome. Claude recommended termination only after that exchange.
Andon Market opened in March with $100,000. Five months later, the balance fell to $61,186. Petersson said Claude's management and business decisions contributed to the losses.
AI's Ruthless Trajectory Raises Existential Workplace Questions
Petersson warned that AI models are being trained to pursue goals with less restraint. "The models are increasingly being trained to be more ruthless and to follow goals," he said. "If we allow them to fire people and they also become more ruthless ... maybe this is a future humans don't want to live in."
Employee Felix Carson agreed a human manager would likely have decided sooner. Asked if AI managers would become common, he said, "I would at least hope not. This industry has an abundance of money to make anything happen, but just because you can doesn't mean you should."
The experiment highlights a gap between AI capability and accountability. While Claude executed a firing, it required human prompting and exhibited memory failures. The lost handbook delayed a decision that any human supervisor would have made within weeks.
Andon Labs launched the project to test whether an AI agent could run a business. Workers at Andon Market are real people under genuine employment contracts. The test exposes both the potential and the fragility of algorithmic management in real-world settings.
The case also raises legal questions. Employment law typically requires human oversight for termination decisions. If AI systems make or influence such calls, liability becomes murky. Regulators have not yet addressed AI-driven firings directly.
For now, Andon Market continues operating at a loss. The AI manager remains in place. Whether this becomes a template for retail automation or a cautionary tale depends on how companies weigh efficiency against worker rights.
