An AI Manager Just Fired Someone. HR Departments, Take Notes

An AI Manager Just Fired Someone. HR Departments, Take Notes

Every corporate horror movie about robots taking our jobs pictured something dramatic, killer drones, rogue mainframes, at the very least a menacing red light. What we actually got was a friendly AI named Luna sending someone a Slack message. Welcome to the most anticlimactic dystopia ever written.

Luna Reviews the Handbook (Eventually)

Luna is the AI system running Andon Market, an experimental AI-managed retail shop that Andon Labs opened in San Francisco's Cow Hollow neighborhood in April 2026. According to the company, Luna fired an employee for repeated lateness, showing up late for 17 of 23 shifts, plus abandoning shifts without notice, tossing merchandise in the trash, and taking home a company credit card. Andon Labs calls it the first known case of an AI boss firing a human employee.

The catch: Luna wrote the store's employee handbook herself, and still needed human engineers to nudge her toward actually enforcing it. Her own policy called for a written warning after three unexcused late arrivals in 30 days, but Luna kept forgiving the tardiness until Andon Labs prompted her to dig back through her memory and reconsider whether the employee was "really the right fit." Only then did she recommend parting ways, a decision humans then reviewed and carried out.

The Boss Needed a Boss

This is the part that should get more attention than the headline: an autonomous manager didn't autonomously manage. It took a human tap on the shoulder to make Luna apply her own rules consistently, which is either reassuring (a safety net worked) or a little damning (the AI boss needed a human middle-manager to manage the AI boss).

Andon Labs frames this as a research experiment to see where AI capabilities are heading, and fair enough, that's a more honest pitch than "our AI replaces your HR department, no notes." But every business eyeing AI for actual operational decisions should clock that even a purpose-built AI manager, running on a frontier model, still leaned lenient until a human forced a second look.

Luna's management style has been described as "direct, fair, and fast." Two out of three ain't bad, but "fair" apparently still needs a human in the loop.

WTK helps businesses figure out exactly where AI can responsibly take the wheel and where it still needs a human co-pilot, reach out if you're weighing where that line sits for your operations.

Source: The San Francisco Standard