An AI agent managing an experimental retail store in San Francisco has made its first firing decision, dismissing a human employee after repeated attendance issues.
Andon Labs, which runs the AI-operated Andon Market, said its AI manager Luna recommended ending the employee’s contract after the worker arrived late for 17 of 23 shifts. The employee had received repeated warnings and additional training over several months before the decision was taken.
The firing was not entirely autonomous. Luna had initially created an attendance policy but later lost track of it, allowing the attendance issues to continue. Andon Labs eventually prompted the AI agent to review its policies and assess whether the employee remained suitable for the role. Luna then recommended parting ways with the employee, with the company reviewing and implementing the decision.
The experiment began when Andon Labs gave Luna a $100,000 budget, internet access and a corporate credit card, with instructions to open a store and make it profitable. Built using Anthropic’s Claude models, Luna has since selected merchandise, hired contractors, posted vacancies, interviewed candidates and hired employees for the store, which sells books, candles, prints, games and branded merchandise.
Human oversight remains in place for legal and ethical issues, while employees are formally employed by Andon Labs and receive guaranteed pay and legal protections. The store has generated sales but is not yet profitable.
The experiment also highlighted a limitation of current AI agents: their tendency to wait for explicit instructions rather than independently identify and act on workplace issues.
