A human barista serves coffee, but an artificial intelligence agent is running operations at an experimental café in Stockholm, where almost all business decisions are handled by AI in a real-world test of automation.

The Andon Café, operated by San Francisco-based startup Andon Labs, has placed an AI agent named “Mona” in charge of management tasks while human staff continue to prepare and serve drinks. Powered by Google’s Gemini, the system oversees hiring, contracts, inventory and day-to-day operations.

International cyberattack disrupts universities and schools across three countries

Since opening in mid-April, the café has earned over $5,700 in sales but remains below break-even, with less than $5,000 left from an initial budget exceeding $21,000. Organisers say most early spending went on setup costs and expect performance to stabilise over time.

Customers have reacted with curiosity, with visitors able to call the AI via an in-café telephone. “It’s nice to see what happens if you push the boundary. The drink was good,” said customer Kajsa Norin.

Andon Labs, founded in 2023, says it is stress-testing AI systems in real environments and has worked with OpenAI, Anthropic, Google DeepMind and xAI. It envisions AI-run organisations in the future.

Technical staff member Hanna Petersson said the café is a controlled experiment to study ethical questions in AI-led management.

However, experts including KTH’s Emrah Karakaya warned of accountability risks, calling it “opening Pandora’s box.”

Mona has made major errors, including over-ordering supplies and mismanaging bread deliveries. Barista Kajetan Grzelczak said jobs remain safe for now, but middle management could be affected.