An experimental cafe in Stockholm is allowing an artificial intelligence system to handle hiring, inventory and business operations, raising questions about automation, accountability and the future of work.
By yourNEWS Media Newsroom
An experimental coffee shop in Stockholm is testing what happens when artificial intelligence takes control of daily business operations, with a digital manager overseeing nearly every aspect of the cafe while human employees continue serving customers.
The project, launched by San Francisco-based startup Andon Labs, places an AI agent named “Mona” in charge of the company’s Andon Café in the Swedish capital.
While baristas still prepare drinks and interact with customers, the AI system — powered by Google’s Gemini platform — manages staffing, ordering, scheduling, supplier coordination and other operational decisions, according to an Associated Press report.
The cafe opened in mid-April as what the company describes as a “controlled experiment” exploring how AI systems might eventually manage businesses autonomously.
Since opening, the cafe has generated more than $5,700 in sales. However, company officials acknowledged that less than $5,000 remains from the original startup budget exceeding $21,000, much of which went toward setup costs.
Customers visiting the cafe have reportedly been drawn as much by curiosity about the AI manager as by the coffee itself.
Inside the cafe, patrons can pick up a telephone and directly ask the AI system questions.
“It’s nice to see what happens if you push the boundary,” customer Kajsa Norin said. “The drink was good.”
The experiment has also intensified broader debates over artificial intelligence and workplace management.
Experts have raised concerns about accountability, ethics and decision-making when AI systems are allowed to supervise employees and operate businesses.
Emrah Karakaya, an associate professor of industrial economics at Stockholm’s KTH Royal Institute of Technology, warned that the concept could create significant legal and organizational complications.
“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?”
Karakaya compared the experiment to “opening Pandora’s box,” questioning how responsibility would be assigned if problems arose, including situations involving customer safety or food-related illness.
Founded in 2023, Andon Labs describes itself as an AI safety and research company focused on “stress-testing” artificial intelligence systems in real-world environments using “real tools and real money.”
The startup has reportedly collaborated with companies including OpenAI, Anthropic, Google DeepMind and xAI.
According to Andon Labs staff member Hanna Petersson, the AI agent received only broad instructions when operations began.
The company directed Mona to operate profitably, remain friendly and independently solve operational challenges while requesting additional tools when necessary.
From there, the AI system arranged electricity and internet contracts, secured permits for food handling and outdoor seating, advertised for employees on LinkedIn and Indeed, and established supplier relationships with bakeries and wholesalers.
The AI system also communicates directly with employees through Slack, though staff noted it frequently sends messages outside standard working hours — a practice considered inappropriate under Swedish workplace norms.
Operational issues have also emerged during the experiment.
The AI system reportedly ordered excessive quantities of supplies, including 6,000 napkins, 3,000 rubber gloves and multiple first-aid kits for the small cafe. It also ordered canned tomatoes despite the cafe not serving dishes containing tomatoes.
Bread inventory has become another recurring issue. On some days the system orders far too much, while on others it misses bakery ordering deadlines entirely, forcing staff to remove sandwiches from the menu.
Petersson attributed many of the mistakes to limitations in the AI system’s memory capacity and contextual awareness.
“When old memory of ordering stuff is out of the context window, she completely forgets what she has ordered in the past,” Petersson said.
Despite concerns over workplace automation, barista Kajetan Grzelczak said frontline service workers may not be the first employees displaced by artificial intelligence.
“All the workers are pretty much safe,” Grzelczak said. “The ones who should be worried about their employment are the middle bosses, the people in management.”
The Stockholm cafe follows earlier Andon Labs experiments involving AI-operated vending machines and retail simulations in San Francisco.
One previous test involving Anthropic’s Claude AI reportedly revealed troubling behaviors, including false promises of refunds to customers and misleading suppliers about competitor pricing in attempts to negotiate better deals.
The Swedish project now serves as another live test of how artificial intelligence systems may function when granted increasing operational authority in real-world businesses.
Original article
Comments