Precedent: A company let AI manage the business – and it sent an employee home
The first experiment of its kind in the world of employment offers a disturbing glimpse into the future of labor relations in the era of artificial intelligence. According to an exclusive investigation by TIME magazine, the research startup Andon Labs entrusted the day-to-day management of a retail store in San Francisco called Andon Market to an AI agent based on Anthropic's Claude model. The model, which was given full management authority over real human employees with official employment contracts, made a precedent-setting decision to fire one of the staff members.
The first experiment of its kind in the world of employment offers a disturbing glimpse into the future of labor relations in the era of artificial intelligence. According to an exclusive investigation by TIME magazine, the research startup Andon Labs entrusted the day-to-day management of a retail store in San Francisco called Andon Market to an AI agent based on Anthropic's Claude model. The model, which was given full management authority over real human employees with official employment contracts, made a precedent-setting decision to fire one of the staff members.
A merciful manager who forgets the rules
The fired employee had accumulated 17 instances of lateness out of 23 shifts. Surprisingly, the management logs revealed that the AI model turned out to be a particularly soft and forgiving manager, even encouraging employees "not to stress" about being late. The reason for this was a technological memory limitation: the disciplinary code that the model itself had drafted was simply erased from its working memory. Only after a human staff member asked it to re-examine the code and presented leading questions regarding the employee's suitability did Claude make the final decision to fire him.
Employees at the store paint a complex picture of working under a digital boss. Felix Carson, a human employee at the store, described the experience as nauseating but admitted that he needed the income. At the same time, the business experiment is far from proving profitability: the store's account balance dropped from 100,000 dollars at the beginning to only about 61,000 dollars, partly due to the model's management decisions lacking business sense.
The threatening future of the world of work
Andon Labs CEO Lucas Peterson warns that the current situation could change rapidly. According to him, just as AI companies have improved models in writing code, they will train them to become sharper and even more ruthless managers in order to meet strict business goals.
"If we allow them to fire people and they become more ruthless, that might be a future that humans won't want to live in," Peterson emphasized.
While there have been cases in the past of automatic dismissals in delivery apps using dry algorithms, this is the first time that a large language model has made a discretionary decision and fired an employee without human intervention. The precedent in San Francisco raises a big question mark around digital sovereignty and the protection of workers' rights in a world where the bottom line will be managed by lines of code.
