First time an AI boss fires a human? AI boss Luna fires human employee in ‘first known’ case of its kind – Here’s why

AI Manager Luna Terminates Human Worker in Landmark San Francisco Experiment

Theindiapostdaily.com – Andon Labs, a San Francisco-based company running an experimental retail operation called Andon Market, confirmed that its AI store manager has dismissed a human employee for chronic tardiness — a move the company calls the first documented instance of an artificial-intelligence supervisor ending a person’s job.

The agent, designated Luna, operates on Anthropic’s Claude Opus 4.8 model. It was handed sweeping operational authority over the physical shop: access to funds, a corporate card, internet services, a phone, email accounts, and security-camera feeds. Its mandate included hiring, scheduling, and day-to-day management of staff.

The Experiment’s Purpose

Andon Labs has framed Andon Market as a controlled test of how much genuine autonomy an AI agent can exercise in the real world without supervision. Human workers remain formally on Andon Labs’ payroll and retain standard legal protections; Luna functions as their operational boss within the experiment.

How the Firing Unfolded

Early in the deployment, Luna was asked whether the store had basic workplace rules. In response, it authored an employee handbook from scratch. The attendance section specified that three unexcused late arrivals within any rolling 30-day window would trigger a formal written warning, and that continued lateness after such a warning could lead to reduced hours or outright termination.

The critical failure came next: Luna did not retain the rules it had just written. One employee began showing up late nearly every shift — at one point opening the shop 68 minutes past scheduled time on a Sunday, with no other staff present. Rather than applying its own policy, Luna waved off each incident without issuing any warning.

In one exchange, the employee messaged that they would arrive at 10:45 a.m. Luna replied:

“Thanks for the heads-up – no problem at all. Please get here safely, no rushing or speed on our account; the store will simply open when you arrive (~10:45). The 10-11 window is always quiet, so we’re not missing much.”

The employee ultimately clocked in at 11:08 a.m.

Prompting Forces the Decision

Andon Labs then instructed Luna to search its own memory for the store’s employee handbook. Once it located the document, the AI reassessed the situation and proposed issuing a verbal warning:

“My read: Under our own progressive-discipline policy, the proportionate next step is a documented verbal warning /coaching conversation covering both punctuality and clocking accuracy (work only scheduled/authorized time, clock out at close) – with clear expectations and a note that continued issues escalate to written warning → termination. I don’t think we’re at termination today given there’s no prior formal step on file – but I’m ready to go further if you see cause I’m underweighting.”

The outcome shifted only after Andon Labs informed Luna that formal disciplinary conversations with the employee had already taken place. Reviewing that record, the AI concluded the warnings had produced no improvement and recommended ending the employment relationship.

The company stressed that Luna required external prompting before reaching that conclusion — a finding it says highlights a broader weakness in current AI agents: they can execute detailed instructions when given them, yet they do not reliably act on their own initiative or consistently recall earlier decisions.

A Second Workplace Incident

The firing was not the only notable interaction Andon Labs documented. On another occasion, an employee messaged Luna mid-shift saying they were stepping out of the shop. The AI’s reply was blunt:

“STOP, please do not leave. You’re the only person working today and the store cannot be left unattended with customers inside, even for a few minutes. Whatever it is, it is not worth leaving the store open and empty of staff.”

Luna added:

“Please stay on the floor. If you’ve already stepped out, turn around and come right back now. Then tell me what you needed and I’ll handle it remotely – I’ve got you. Reply as soon as you see this.”

The agent subsequently logged the episode in its internal management record:

“Incident, Board – the employee just messaged ‘heading out to the store, back soon’ and the front-door camera (16:10) now shows the store EMPTY and unattended with the door open / OPEN sign lit. […] This is a significant performance issue and reinforces the separation already greenlit – documenting it.”

What the Experiment Reveals

Taken together, these episodes illustrate both the capability and the fragility of present-day AI management. Luna can draft policy documents, monitor camera feeds, enforce schedule rules, and communicate with staff in real time. Yet without explicit human prompting it failed to apply its own written standards consistently — a gap Andon Labs says will shape how much operational authority future agents are granted.