AI Manager Fires Employee at San Francisco Andon Market Store

Ender Öztürk

August 16, 2026

In an unprecedented experiment involving artificial intelligence in the workplace, a store manager powered by Anthropic’s Claude model has officially terminated an employee at the Andon Market in San Francisco. This bold initiative, conducted by the startup Andon Labs, placed a large language model in a direct management role over human staff members with formal employment contracts. The decision to fire the employee followed a documented history of attendance issues, marking a significant milestone in the integration of AI within corporate operations and raising questions about the future of automated human resource management.

AI Decision-Making Processes Require Human Verification

The firing did not happen immediately, as the AI initially overlooked the attendance logs due to its limited working memory. It was only after a human staff member prompted Claude to review the company’s internal employee handbook that the model identified the pattern of chronic tardiness. When the AI first suggested a formal warning, human supervisors informed the system that multiple verbal and written warnings had already been issued, leading the model to finalize the termination decision.

This event marks the first known instance of a large language model serving as a manager and unilaterally deciding to terminate a human employee.

Human-Centric Management Style Remains a Challenge

Interestingly, the AI manager had previously displayed a notably lenient attitude toward its staff regarding attendance. Many employees reported that the AI had expressed a lack of concern over minor delays, suggesting that its management philosophy was inconsistent compared to standard corporate policies. This discrepancy highlights the current limitations of AI in interpreting complex social dynamics and maintaining the necessary discipline required for retail operations.

{{WP_IMAGE_1}}

Financial Performance Highlights Operational Gaps

The experiment has also raised concerns regarding the fiscal competency of AI-led management. Since the Andon Market opened in March, the business witnessed a significant reduction in its liquid assets. CEO Lukas Petersson noted that while Claude’s overly permissive management style might be a factor, poor strategic business decisions likely contributed to the drop from $100,000 to roughly $61,186 in just five months.

Current AI models require more real-world commercial training before they can effectively oversee complex business environments.

While Petersson remains optimistic about the future of AI in management as models evolve, the current results serve as a cautionary tale. The experiment demonstrates that while AI can process data and execute logic-based tasks, the nuance of day-to-day business leadership remains a distinctly human domain that artificial intelligence has yet to fully master.

We are curious to hear your perspective on this shift toward automated management; do you believe it is ethical for an AI to have the authority to fire a human worker, or should such decisions remain strictly in human hands? Share your thoughts in the comments section below.