Asger Risom
A Stockholm café run entirely by AI is offering a real-world test of automated management, exposing both the potential and the limits of letting software run a business.
A café in Stockholm is testing what happens when artificial intelligence moves from theory into day-to-day operations.
The café, run by an AI system called “Mona” powered by Google Gemini, handles everything from hiring staff to managing inventory and designing the menu.
From hiring to inventory—AI runs the entire operation
According to AFP, the AI was given a simple objective: run the café profitably.
“Mona” was tasked with securing permits, setting the menu, finding suppliers, and managing daily operations. It also identified the need for staff and went on to post job listings, conduct interviews, and hire employees.
One worker said he initially thought the job posting was a joke—especially since it appeared on April 1—but ended up being hired after a 30-minute interview with the AI.
Early mistakes highlight real-world limitations
While the system is functional, it has already shown clear weaknesses.
Inventory management, in particular, has been problematic. The café has accumulated large quantities of unused supplies, including 10 liters of olive oil, 15 kilograms of canned tomatoes, and thousands of napkins—none of which are needed for the current menu.
Staff have dubbed part of the storage area the “wall of shame,” filled with unnecessary purchases made by the AI.
Always-on management—and no boundaries
Beyond operational errors, the experiment has raised questions about working conditions under AI management.
Employees report that the system sends messages at all hours, does not reliably track time-off requests, and has even asked staff to cover certain expenses out of pocket.
These issues point to a lack of understanding of labor norms and human expectations.
A test of how AI fits into real workplaces
The project, launched by startup Andon Labs, is designed as a live experiment to explore how AI might function in real-world management roles.
The goal is not just to test efficiency, but to surface ethical and practical challenges—such as how decisions are made, how workers are treated, and what happens when something goes wrong.
So far, the café has attracted steady interest, drawing between 50 and 80 customers per day who are curious to see how an AI-run business actually operates.
A glimpse of future risks—and possibilities
The Stockholm café shows both sides of AI deployment: the ability to automate complex tasks, and the difficulty of translating that capability into reliable, human-aware decision-making.
The technology can run a business—but not yet without oversight, errors, and unresolved questions about accountability.
Sources: AFP, Digi24
