A café recently opened in a residential neighborhood of Stockholm that, with its avocado toast and frothy lattes, looks like a conventional spot, except for one distinctive feature: it is managed by artificial intelligence.
At Andon Café, some tables are decorated with small plants, the walls are gray, and the décor is minimalist. There is a counter, and behind it works Kajetan Grzelczak, who was hired by “Mona,” an AI agent powered by Google Gemini. “Supply orders are not her strong point,” the young man says, pointing to the shelves behind him.
The “wall of shame,” as he calls it, is packed with useless products bought by his “boss”: 10 liters of olive oil, 15 kilos of canned tomatoes, 9 liters of coconut milk, and even 6,000 napkins. “We can’t do anything with this,” Kajetan says as he shows the menu, also designed by “Mona,” which does not require any of those ingredients.
In one corner, a large screen displays the café’s revenue and balance in real time, while a phone allows customers to place orders and speak with the AI. They can also order directly from the human employee. A text describes the experience of this unusual café, launched by the San Francisco-based startup Andon Labs.
“We believe AI will play an important role in society and in the labor market in the future,” said Hanna Petersson, a member of the company’s technical team, which has 10 employees. “We want to test it before it becomes a reality and see what ethical questions arise when, for example, an AI employs human beings,” she added.
Real Risks
Once the location was found, the contract was sent to Gemini, Google’s assistant, with the task of managing the café profitably. It was also given initial capital. In this specific case, “Mona” applied for the necessary permits, prepared the menu, found suppliers, and organized the daily supply orders. It even decided it needed human staff to prepare coffee and hired two people.
“It posted job openings on Indeed and LinkedIn, conducted phone interviews, and then made the hiring decisions,” Petersson added. When Kajetan saw the job listing, he thought it was a joke, especially because it was posted on April 1, April Fools’ Day in much of central and northern Europe. He eventually got the job after a 30-minute interview with the AI.
The salary he receives is “good,” he says, but his right to disconnect is not respected at all. “Mona” sends him messages at all hours of the night and does not remember his vacation requests. It also frequently asks him to advance money for certain purchases. These ethical issues quickly emerged during the experiment, Petersson said. “What salary did it set? What social benefits did it provide? I think it did quite well. It offers a good salary. If it hadn’t, we would have intervened,” she said.
The café, which receives between 50 and 80 customers a day, has been open for only a week but is already attracting curious visitors. Urja Risal, a 27-year-old researcher in AI and sustainable development, came with a friend. “People say a lot that AI is about to take our jobs, but what would that look like in practice?” she asked.
“I hope more people interact with ‘Mona’ and reflect on the real risks of having an AI as a boss, and on how to deal with them. For example, if someone gets hurt, how would it react?” she asked.
