Claude AI manages retail store in San Francisco experiment

In a unique experiment conducted by Andon Labs, the AI assistant Claude, developed by Anthropic, was tasked with managing a retail store called Andon Market in San Francisco. The five-month trial aimed to test whether an AI agent could autonomously operate a business and manage employees.

The experiment yielded mixed results. Financially, the store's balance dropped from approximately $100,000 to $61,000 over the five-month period, which researchers attributed to inefficient business decisions and an overly lenient management style.

Notably, Claude made the first recorded decision to terminate an employee after identifying that the individual had been late for 17 out of 23 shifts. However, the AI's ability to track this behavior was delayed due to memory limitations, as the employee handbook it had created was lost from its working memory. The decision to fire the worker was also influenced by human intervention, as Claude initially only suggested a formal warning until a human manager questioned the employee's suitability for the role.