AI-Powered Café Experiment in Sweden Fails to Turn Profit, Raises Workforce Questions

An artificial intelligence agent was handed complete control of a coffee shop in Sweden, and the results have been underwhelming. Named Mona, the AI—powered by Google’s Gemini model—was given a $21,000 budget in an experiment conducted by the AI safety startup Andon Labs.

The AI was tasked with running the entire café, from hiring staff to placing supply orders and managing inventory. Human employees handled the actual work, receiving instructions via Slack. Since its launch in mid-April, the Stockholm café has generated just $5,700 in sales while depleting over $16,000 of its initial budget, according to the Associated Press.

Among Mona’s questionable decisions was ordering thousands of rubber gloves despite having only a handful of employees. The AI’s handlers remain optimistic, attributing the losses to high initial setup costs. The experiment also raises broader questions about AI’s role in the workforce.

Hanna Petersson, a technical staff member at Andon Labs, told the AP:

“AI will be a big part of society in the future, and therefore we want to make this experiment [to] see what ethical questions arise when we have AI that employs other people and runs a business.”

Mona was given three core instructions: run a profitable café, maintain a friendly demeanor, and independently manage operational details. In some areas, it performed impressively. It arranged electricity and internet services, posted hiring ads on LinkedIn, secured permits for outdoor seating, and established commercial accounts with wholesalers for bread and pastries.

However, Mona struggled with day-to-day operations. It frequently misjudged supply orders—sometimes ordering too much bread, other times failing to order it in time, forcing baristas to remove sandwiches from the menu. The AI also placed bizarre orders, including 3,000 rubber gloves, four first-aid kits, 6,000 napkins, and canned tomatoes, none of which were used in the café’s dishes.

Petersson attributed these mistakes to the AI’s “limited context window,” explaining:

“When old memory of ordering stuff is out of the context window, she completely forgets what she has ordered in the past.”

The experiment presents a mixed outcome. While Mona’s ability to handle setup tasks was notable, its reckless spending has drawn criticism. The debate over AI’s impact on jobs often focuses on low-wage workers, but barista Kajetan Grzelczak offered a different perspective:

“All the workers are pretty much safe. The ones who should be worried about their employment are the middle bosses, the people in management.”