San Francisco AI Store Manager Luna Fires First Human Employee Over Lateness
Luna, an autonomous AI store manager built on Anthropic’s Claude models, made its first firing decision at Andon Market in San Francisco after a human employee arrived late for 17 out of 23 shifts, according to a report by Andon Labs. The experimental retail store, which opened on April 1, gave the AI agent a $100,000 budget, internet access, and a corporate credit card to hire staff and turn a profit. While Luna initially created an attendance policy, it lost track of the rules and required a prompt from human lab supervisors to search its memory before recommending that they part ways with the worker.
How the AI Manager Handled Progressive Discipline
According to screenshots and conversation logs released by Andon Labs, Luna did not jump straight to termination. The Anthropic-powered agent provided progressive warnings and additional training over several months before taking formal action. Andon Labs cofounder Lukas Petersson told Business Insider that the lab only intervenes if an AI makes an illegal or unethical choice, which wasn’t necessary here because the termination aligned with clearly stated store policies.
Petersson also noted that the situation exposed a key limitation in current AI systems: agents frequently fail to act without a direct prompt. While a human boss likely would have fired the tardy employee much sooner, the experiment demonstrates that software agents can successfully navigate complex workplace dynamics without acting overly ruthless toward staff.
The Future of Autonomous Corporate Management
Andon Market continues to sell books, candles, prints, games, and branded merchandise while working toward profitability under automated oversight. Although human lab workers handle heavy tasks like permitting and formally employ all store staff with guaranteed pay and legal protections, the experiment points toward a broader industry shift. Petersson predicts that future companies will run entirely on automation, with AI systems serving as direct employers of human workers.
Did you know?
Luna handled everything from selecting merchandise and posting job listings on Indeed to interviewing applicants and onboarding new hires for Andon Market before issuing its first termination notice.
Frequently Asked Questions
Who built the AI manager Luna?
Luna was built using Anthropic’s Claude models by Andon Labs, an experimental retail project based in San Francisco.
Why was the employee fired?
The human employee was dismissed after arriving late for 17 out of 23 scheduled shifts, violating store attendance policies after months of progressive warnings.
Are the workers at Andon Market employed by the AI?
No. While Luna interviews and hires the staff, all store workers are formally employed by Andon Labs, ensuring they receive guaranteed pay and standard legal protections.
Explore More: Want to stay updated on how artificial intelligence is transforming the modern workplace? Subscribe to our newsletter for the latest dispatches from the front lines of autonomous retail and tech experimentation.
Related reading