Srinagar, Aug 17: An artificial intelligence system powered by Anthropic’s Claude has reportedly fired a human employee at an experimental retail store in San Francisco, marking a striking development in the evolving role of AI in workplace management.
The Indian Express, citing a report by Time, reported that “Luna”, an AI agent deployed by AI research startup Andon Labs to manage its experimental Andon Market store, decided to terminate a human employee after repeated attendance violations.
According to the report, the employee had arrived late for 17 of 23 scheduled shifts. Luna had earlier drafted an employee handbook for the store, but the relevant rules subsequently dropped out of its working memory, delaying any action on the employee’s conduct.
The decision, however, was not entirely autonomous. Store management logs cited by Time showed that a human member of Andon Labs prompted Luna to reconsider the employee’s record after the AI initially recommended a formal warning rather than dismissal. Following further prompting and additional context about the employee’s repeated lateness, Luna ultimately recommended termination.
The episode is significant because Luna is not merely being used to automate an individual workplace task. Andon Labs has given the AI responsibility for running a real retail business, including selecting products, setting prices, managing operations and hiring human staff. The company says Luna has access to a corporate card, internet, email, a phone number and security-camera feeds.
Andon Market, located in San Francisco’s Cow Hollow neighbourhood, was established as a controlled experiment to examine how AI agents behave when entrusted with real-world commercial responsibilities. The company gave Luna a three-year retail lease and a substantial operating budget, while human employees remain formally employed by Andon Labs and retain legal and employment protections.
Andon Labs co-founder Lukas Petersson told Time that the dismissal did not necessarily represent excessive strictness by the AI, arguing that a human manager might have taken the same action much earlier. At the same time, the episode has highlighted a more consequential question: whether AI systems should eventually be entrusted with decisions that directly affect people’s employment and livelihoods.
The experiment also exposes the limits of current AI autonomy. Although Luna was capable of identifying the attendance problem and ultimately recommending dismissal, the decision was reached only after human intervention and prompting. That distinction is crucial as companies increasingly explore AI agents capable of acting with greater independence.
The development has consequently intensified debate over accountability, transparency and human oversight in AI-driven management. While AI systems are rapidly moving from assisting workers to supervising processes, the San Francisco experiment offers an unusually concrete glimpse of what could happen when an AI is placed not merely alongside employees, but above them in the chain of command.
For now, Luna’s authority remains part of a controlled experiment. But its first reported dismissal of a human worker represents a notable milestone in the transition from AI as a workplace tool to AI as a workplace decision-maker.
