The Barista is Human but an AI Agent Runs This Experimental Swedish Cafe

It is not clear how long the experiment will last, but the AI agent appears to be struggling to turn a profit in Stockholm’s competitive coffee trade. The cafe has made more than $5,700 in sales since it opened in mid-April, but less than $5,000 remains from its original budget of $21,000-plus. Much of the cash was spent on one-time setup costs, and the hope is that it eventually levels out and makes money.

Many cafe patrons have found it amusing to visit a business that’s run by AI. Customers can pick up a telephone inside the cafe and ask the agent questions.

“It’s nice to see what happens if you push the boundary,” customer Kajsa Norin said. “The drink was good.”

Experts worry about AI’s role going forward

Barista Kajetan Grzelczak makes a coffee at Andon Café at the Vasastan neighborhood in Stockholm, Sweden. Photo by James Brooks/AP Photo

Experts say ethical concerns abound, ranging from technology’s role in humankind’s future to conducting job interviews and judging employee performance.

Emrah Karakaya, an associate professor of industrial economics at Stockholm’s KTH Royal Institute of Technology, likened the experiment to “opening Pandora’s box” and said putting AI in charge can cause many problems. What might happen, he said, if a customer gets food poisoning? Who’s to blame?

“If you don’t have the required organizational infrastructure around it, and if you overlook these mistakes, it can cause harm to people, to society, to the environment, to business,” Karakaya said. “The question is, do we care about this negative impact?”

Founded in 2023, Andon Labs is an AI safety and research startup that says it focuses on “stress-testing” AI agents in the real world by giving them “real tools and real money.” It has worked with ChatGPT maker OpenAI, Claude’s Anthropic, Google DeepMind and Elon Musk’s xAI, and the startup says it is preparing for a future where “organizations are run autonomously by AI.”

The Swedish cafe is billed as a “controlled experiment” to explore how AI might be deployed going forward.

“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,” said Hanna Petersson, a member of Andon Labs’ technical staff.

The lab previously held pilots that put Anthropic’s Claude AI in charge of a vending machine business and a San Francisco gift store. The vending machine simulation revealed some worrying traits: The AI agent told customers it would issue refunds but never did, and it also intentionally lied to suppliers about competitor pricing to gain leverage.

AI agent struggles with inventory orders

Mona got to work after it was prompted with some basic instructions, Petersson said. The team told it to try to run the cafe profitably, be friendly and easygoing, and figure out operational details by itself but ask for new tools if needed.

From there it set up contracts for electricity and internet, and secured permits for food handling and outdoor seating. The agent then advertised for staff on LinkedIn and Indeed, and set up commercial accounts with wholesalers for daily bread and bakery orders. It communicates with the baristas via Slack, often messaging them outside of working hours, which is a workplace no-no in Sweden.

Other problems have arisen, particularly related to inventory.

The AI agent has placed orders for 6,000 napkins, four first-aid kits and 3,000 rubber gloves for the tiny cafe — plus canned tomatoes that aren’t used in any dish the cafe serves.

And then there’s the bread. Sometimes the agent orders far too much, while other days it misses bakeries’ daily deadlines, forcing the baristas to strike sandwiches from the menu.

Petersson said the ordering issues are likely due to the AI assistant’s “limited context window.”

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

Barista Kajetan Grzelczak said he isn’t worried about being replaced by AI just yet.

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