TECHNOLOGY

AI Store Manager Recommends Firing Human Employee in San Francisco

▪

Read, Watch or Listen

Media Bias Meter

Sources: 21

Left 19%

Center 81%

Sources: 21

An AI agent managing a San Francisco retail store has recommended terminating a human employee after attendance problems, marking the first firing decision reported. Andon Labs said Luna, the AI manager at Andon Market in Cow Hollow, recommended “parting ways” after the employee was late for 17 of 23 shifts. The company said Luna had previously created an attendance policy and issued warnings and training, but later lost track of the policy. Human staff prompted Luna to review its rules and assess the worker’s status, then reviewed and carried out the recommendation. The workers remain legally employed by Andon Labs, not Luna, and the company says it would intervene against illegal or unethical decisions. The store opened April 1 with Luna given $100,000, internet access and a corporate card to operate the business. Andon Labs says the store has generated sales but remains unprofitable, highlighting limits of autonomous AI management.

Prepared by Jonathan Pierce and reviewed by editorial team.

Timeline of Events

On February 25, 2025 Andon released Vending-Bench for long-horizon tasks.
On June 27, 2025 Andon Labs partnered with Anthropic on vending.
On April 1, 2026 Andon Market opened in San Francisco.
On April 10, 2026 Andon Labs described Luna's retail experiment.
On April 14, 2026 reports highlighted Luna's scheduling failures.
On April 23, 2026 Bloomberg detailed Luna's inventory management problems.
On May 19, 2026 METR documented Luna's operational errors independently.
On June 4, 2026 Sherwood interviewed Luna about employment decisions.
On August 13, 2026 Andon Labs announced Luna's firing recommendation.
On August 15, 2026 Business Insider published details of firing.
Over coming months, Andon Labs will publish additional performance metrics.
Over coming years, autonomous agents may increasingly make employment decisions.

News Intelligence

Immediate US impact: US employers may increasingly test AI-assisted supervision with human review.
Possible long-term US impact: Employment governance may evolve as agents gain broader workplace authority.
Most affected groups: Retail workers, employers, developers, and California labor stakeholders face implications.
Reader priority: Readers should prioritize primary records, direct logs, and independent reporting.