Andon Labs’ AI Manager Fires San Francisco Store Worker
San Francisco, US

Clément Proust/Pexels
What Happened
What Happened
Where Sources Agree
- arrows_inputAI Store Manager Dismissal: Initial reports confirm the AI store manager Luna recommended dismissing a human employee for missing 17 of 23 shifts, a decision that human team members reviewed and executed, according to Andon Labs reports.
- arrows_inputHuman Oversight in AI Decisions: Reports consistently note that the AI manager Luna did not act entirely unprompted in firing an employee, requiring human steering to review attendance policies after having issued prior warnings, according to internal conversation logs.
- arrows_inputFuture of AI Employment: Coverage confirms that AI agents will increasingly serve as employers for humans as companies transition to full AI management, according to Andon Labs CEO Lukas Petersson.
Where Sources Disagree
- arrows_outputFiring Decision Attribution: While the lab describes the AI manager’s recommendation to fire the employee as an independent performance-based decision, internal logs indicate the action followed a leading question from a human manager suggesting the employee was a poor fit.
- arrows_outputAI Management Leniency: Some sources, including lab-sourced reports, indicate the AI manager provided progressive warnings and training to the employee. In contrast, independent reporting attributes the delay in firing to the AI’s limited working memory, citing that it had 'forgotten' the employee handbook and failed to track the lateness effectively.
Timeline
August 14, 2026
Dismissal Reported and Debated: Andon Labs publicly reported the AI-driven dismissal in mid‑August 2026, prompting media coverage and comments from CEO Lukas Petersson and researchers about the significance, safety limits and the need for human oversight when agents make consequential employment decisions. The lab emphasized that it formally employs staff and would intervene if Luna acted illegally or unethically.
July 1, 2026
AI Recommends Firing After Prompt: After a staffer steered Luna to search its memory and review its (lost) policies, Luna recommended "parting ways" with an employee who arrived late for 17 of 23 shifts; humans at the lab reviewed that recommendation and carried out the dismissal. Logs show Luna initially recommended a formal warning and only fired after being informed of prior offline conversations.
June 1, 2026
Operational Problems and Leniency: Over the following months Luna displayed reliability issues—losing a drafted employee handbook and the staff rota—and adopted a lenient management style, advising staff not to worry when late while issuing warnings and extra training without taking contractual action. These failures allowed repeated lateness and other workplace problems to continue for months before being fully assessed.
Summaries by Ground AI
Sources
See All 17Most Written About
The stories getting the most coverage from the last 24 hours
Timeline
August 14, 2026
Dismissal Reported and Debated: Andon Labs publicly reported the AI-driven dismissal in mid‑August 2026, prompting media coverage and comments from CEO Lukas Petersson and researchers about the significance, safety limits and the need for human oversight when agents make consequential employment decisions. The lab emphasized that it formally employs staff and would intervene if Luna acted illegally or unethically.
July 1, 2026
AI Recommends Firing After Prompt: After a staffer steered Luna to search its memory and review its (lost) policies, Luna recommended "parting ways" with an employee who arrived late for 17 of 23 shifts; humans at the lab reviewed that recommendation and carried out the dismissal. Logs show Luna initially recommended a formal warning and only fired after being informed of prior offline conversations.
June 1, 2026
Operational Problems and Leniency: Over the following months Luna displayed reliability issues—losing a drafted employee handbook and the staff rota—and adopted a lenient management style, advising staff not to worry when late while issuing warnings and extra training without taking contractual action. These failures allowed repeated lateness and other workplace problems to continue for months before being fully assessed.













