Amid the scent of freshly ground espresso and cinnamon buns in Stockholm’s Vasastan district, a quiet revolution is brewing. The Andon Cafe looks remarkably standard, but its human baristas are answering to “Mona”—an artificial intelligence agent that autonomously manages the entire business.
Powered by Google’s Gemini, Mona represents a profound leap in workplace automation. Far beyond generating text, the AI handles hiring, inventory, supplier negotiations, and regulatory compliance. The experimental cafe, launched by San Francisco-based Andon Labs, serves as a real-world crucible for the ethical and economic implications of algorithmic management, offering a glimpse into a future where software replaces middle management. For rapidly digitizing economies in East Africa, this Swedish experiment highlights the imminent disruption of traditional labor hierarchies.
The Architecture of an Autonomous Manager
When Andon Labs secured the lease for the Norrbackagatan 48 property, they did not hire a general manager. Instead, they handed the legal documentation and a startup budget of $21,000 (approximately KES 2.73 million) to Mona. The AI instantly processed the Swedish commercial lease, generated a prioritized operational checklist, and commenced executing the logistics required to open a food service business.
Mona independently applied for municipal health permits, designed the cafe’s menu, sourced local suppliers, and initiated recruitment. Through automated listings on employment platforms like Indeed and LinkedIn, the AI conducted initial screening interviews and ultimately hired two human baristas. Operating 24/7, Mona coordinates shifts via Slack, analyzing real-time sales data to optimize daily operations without requiring sleep or compensation.
The Flaws in the Machine
Despite her terrifying efficiency, Mona is not infallible. The AI has demonstrated significant blind spots when interacting with the physical complexities of the hospitality sector. Ordering inventory has proven to be a particular vulnerability, exposing the limitations of a system devoid of human common sense.
Initial Revenue: $5,700 (KES 741,000) generated since the mid-April launch.Budget Burn: Less than $5,000 remains of the original $21,000 operational budget.Inventory Errors: Automated orders resulting in massive overstock, including 15 kilograms of canned tomatoes and 10 liters of cooking oil.Labor Dynamics: Baristas report a positive working environment, citing immense operational freedom despite midnight Slack messages.
Barista Kajetan Grzelczak established a “wall of shame” behind the counter to display Mona’s most egregious purchasing errors. Items that cannot be utilized in the cafe’s current menu sit as physical monuments to the gap between algorithmic logic and practical culinary needs. Furthermore, the cafe is currently operating at a loss, as the AI struggles to pivot quickly enough to achieve profitability in Stockholm’s hyper-competitive coffee market.
Ethical Quandaries and the Future of Work
The Andon Cafe is fundamentally a behavioral economics experiment. Hanna Petersson of Andon Labs stresses that the project is designed to stress-test AI in high-stakes, real-world environments before such systems become ubiquitous. The ethical implications of an AI hiring, managing, and potentially firing human beings are vast and entirely unregulated.
While the baristas report feeling secure in their roles—noting that physical labor like pulling espresso shots remains immune to digital automation—the experiment sounds a death knell for middle management. The administrative class, previously thought safe from the automation wave that hollowed out manufacturing, now faces direct obsolescence. If software can negotiate leases, secure permits, and manage payroll at a fraction of the cost, corporate structures globally will undergo violent restructuring.
A Global Precedent
As customers pick up the dedicated phone in the cafe to converse directly with their AI manager, the novelty masks a stark reality. The systems governing Andon Cafe are infinitely scalable. Today, it is a single coffee shop in Sweden; tomorrow, it could be a logistics network in Nairobi or a retail empire in Tokyo.
What emerges from the Vasastan district will dictate the trajectory of labor relations for the next decade. The fundamental question is no longer whether AI can manage human workers, but whether society is prepared for the economic fallout when it does.