In Stockholm, Andon Cafe operates under the direction of an AI agent called “Mona,” which handles decisions ranging from hiring baristas to managing suppliers an experiment that raises substantive questions about the limits of automation in the workplace.
An AI agent as a real business manager
Andon cafe, located in Stockholm, is operated almost entirely by an artificial intelligence system called “Mona,” developed by Andon Labs using advanced language models. The system was responsible for applying for business licenses, identifying suppliers, designing the menu, and hiring the staff responsible for in-person service.
The project is designed to investigate the feasibility of AI agents as autonomous managers in real business environments.
A hiring process conducted without human involvement
One of the most notable aspects of the experiment is recruitment. Mona independently posted job listings, conducted interviews, and made hiring decisions. During the process, the system rejected applicants with advanced academic credentials due to a lack of practical experience in the service industry — prioritizing hands-on competence over formal qualifications. Two baristas were ultimately selected to handle day-to-day operations.
Operational limitations identified during the experiment
Despite demonstrating capability across several management tasks, the experiment also exposed significant lapses in judgment. The system placed excessive supply orders, including quantities incompatible with the café’s operational capacity. It also assigned tasks at unreasonable hours and instructed employees to make purchases using their personal funds pointing to meaningful gaps in practical reasoning and financial management.
A controlled experiment with an academic purpose
Andon Labs clarifies that this is a controlled research project. All workers are employed directly by the research company, and none of them rely solely on the AI’s decisions for employment purposes. The stated goal is to observe the behavior of autonomous systems in real-world settings and inform discussions about the ethical and practical parameters of their deployment.
Implications for the future of work
The case raises questions that extend well beyond the café setting: to what extent can autonomous systems be trusted in management roles? What oversight mechanisms are necessary when AI makes decisions that affect people? And how should accountability be distributed when errors occur?
Though still experimental, the project illustrates a broader shift underway: artificial intelligence is moving from a support tool to a decision-making agent with all the implications that entails.
