AI Manager Luna Fires Human Worker at San Francisco Store in Real-World Management Experiment

An artificial intelligence system powered by Anthropic’s Claude has taken the unusual step of recommending the dismissal of a human worker at an experimental retail store in San Francisco.

The AI manager, known as Luna, runs Andon Market, a real-world retail experiment created by AI research company Andon Labs. Luna has been given responsibility for several business operations, including hiring staff, choosing products, setting prices and managing day-to-day activities. Click Here To Follow Our WhatsApp Channel

The latest development has attracted attention because it moves beyond AI simply performing individual tasks. In this experiment, an AI agent is being given responsibility for decisions that directly affect human employees.

AI Manager Recommends Employee’s Dismissal

According to reports, Luna decided that one of the store’s workers should be dismissed after repeated attendance problems.

The employee had reportedly arrived late for 17 of 23 shifts, despite receiving warnings and training.

Luna recommended ending the employee’s employment. However, the decision was not completely autonomous. Human researchers at Andon Labs reviewed the situation and prompted the AI to reconsider the employee’s performance before the recommendation was ultimately approved and carried out.

This detail is important because it shows that, despite Luna’s broad responsibilities, humans remain involved in consequential employment decisions.

What Is Andon Market?

Andon Market opened in San Francisco in April 2026 as an experiment to examine how an AI agent behaves when given responsibility for running a real business.

Andon Labs gave Luna a three-year retail lease, a budget of about $100,000, internet access and a corporate credit card. The AI was given a broad objective: operate the store and make a profit.

Luna has been involved in many aspects of the business.

The AI has selected merchandise, dealt with suppliers, helped design the store, recruited employees and managed daily operations. It even posted job advertisements and conducted interviews when looking for staff.

The experiment is designed to test what happens when AI agents are given greater autonomy in the real world rather than being restricted to a computer screen.

Humans Still Have Legal Responsibility

Despite Luna’s managerial role, the human employees are not legally employed by the AI.

Andon Labs has said that the workers are formally employed by the company and receive wages and legal protections. The company has also stressed that no employee’s livelihood should depend solely on an AI system’s judgement.

This human oversight is particularly important when AI systems make decisions involving employment, pay or dismissal.

An AI can process information quickly, but it may not understand the full context behind a person’s behaviour. Attendance problems, for example, can sometimes involve circumstances that are not immediately visible in workplace data.

A New Question for the Workplace

The Luna experiment raises a much bigger question: Could AI managers eventually become responsible for human workers?

Businesses already use AI to assist with recruitment, scheduling, performance analysis and other workplace tasks.

Luna represents a more advanced model in which one AI agent has been given responsibility for bringing many of those functions together.

The experiment therefore offers an early look at a possible future in which AI systems could manage teams, make operational decisions and recommend action against employees.

But it also highlights the risks.

Who Is Accountable When AI Makes a Bad Decision?

One of the biggest challenges is accountability.

If an AI manager makes an unfair or incorrect decision, who is responsible?

The company?

The developers who built the AI?

The people who gave it authority?

Or the human managers who approved its decision?

These questions become increasingly important as AI agents move from assisting humans to making decisions with real-world consequences.

Luna’s dismissal decision demonstrates why human oversight remains essential. Even in an experiment specifically designed to test AI autonomy, researchers still had to intervene before the employment decision was finalised.

AI Is Not Yet Ready to Replace Human Managers

The experiment does not prove that AI is ready to replace human managers.

In fact, some of Luna’s earlier experiences have demonstrated the limitations of autonomous AI systems. Andon Labs has used real-world experiments to identify situations in which AI agents make mistakes, misunderstand instructions or require human intervention.

That is precisely why the San Francisco store is significant.

It is less a demonstration that AI has solved management and more a test of what happens when AI is given genuine responsibility.

The results could help researchers understand where AI agents perform well, where they fail and what safeguards are needed before such systems are used more widely.

The Future of AI Management

Luna’s decision to recommend firing a human employee may sound like a scene from a science-fiction film, but the experiment represents a real development in the evolution of AI agents.

The technology is moving beyond answering questions and generating content towards planning, decision-making and taking actions in the real world.

That could eventually transform how businesses operate.

But greater autonomy must come with greater accountability.

AI systems may become powerful management tools, but important decisions affecting people’s jobs should continue to involve responsible human oversight until the technology can demonstrate that it can handle such responsibilities safely, fairly and reliably.

The story of Luna is therefore not simply about an AI firing a human worker.

It is an early warning — and an experiment — about what happens when artificial intelligence moves from being an employee’s tool to becoming the boss.