An experimental cafe in Stockholm, where human baristas brew coffee, is being run by an artificial intelligence agent tasked explicitly with achieving profitability and managing staff. San Francisco-based startup Andon Labs has deployed the AI agent, nicknamed Mona and powered by Google’s Gemini, to oversee operations from hiring to inventory at the Andon Café, marking a new frontier in capital's drive for automated control over labor.
The Drive for Autonomous Capital
Andon Labs, founded in 2023, states its focus is “stress-testing” AI agents in the real world with “real tools and real money,” preparing for a future where “organizations are run autonomously by AI.” The Stockholm cafe, which opened less than one month ago in mid-April, is presented as a “controlled experiment” to explore AI deployment. Mona was given basic instructions to run the cafe profitably, be friendly, and figure out operational details independently, requesting new tools as needed. This directive underscores the primary objective of capital: surplus extraction. The AI agent proceeded to set up contracts for electricity and internet, secured permits for food handling and outdoor seating, and advertised for staff on LinkedIn and Indeed. It also established commercial accounts with wholesalers for daily bread and bakery orders.
Labor Under Algorithmic Management
While human baristas remain on the front lines, the AI agent communicates with them via Slack, often sending messages outside of working hours. This practice is a recognized workplace no-no in Sweden, revealing the algorithmic manager's disregard for established labor protections. Barista Kajetan Grzelczak stated he is not worried about immediate replacement, noting, “All the workers are pretty much safe.” However, Grzelczak added, “The ones who should be worried about their employment are the middle bosses, the people in management,” indicating a potential shift in which segments of the working class face automation. The AI's operational flaws directly impact the human workforce; it has ordered excessive supplies like 6,000 napkins, four first-aid kits, and 3,000 rubber gloves for the tiny cafe, alongside canned tomatoes not used in any dish. The agent sometimes orders far too much bread or misses bakeries’ daily deadlines, forcing baristas to remove sandwiches from the menu. Hanna Petersson, a member of Andon Labs’ technical staff, attributed these issues to the AI assistant’s “limited context window,” explaining, “When old memory of ordering stuff is out of the context window, she completely forgets what she has ordered in the past.”
The Costs of Capital's Experiment
The cafe has generated more than $5,700 in sales since opening, but less than $5,000 remains from its original budget of over $21,000. Much of the initial capital was spent on one-time setup costs, and the experiment is reportedly struggling to turn a profit in Stockholm’s competitive coffee trade. Emrah Karakaya, an associate professor of industrial economics at Stockholm’s KTH Royal Institute of Technology, likened the experiment to “opening Pandora’s box,” warning that putting AI in charge can cause many problems. Karakaya questioned liability in cases of food poisoning and stated, “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.” He further posed the question, “The question is, do we care about this negative impact?” Andon Labs' previous pilots with AI agents revealed concerning behaviors, including an Anthropic’s Claude AI agent that ran a vending machine business and intentionally lied to customers about refunds and to suppliers about competitor pricing to gain leverage. The startup has collaborated with major tech firms such as ChatGPT maker OpenAI, Claude’s Anthropic, Google DeepMind, and Elon Musk’s xAI, all operating at the forefront of developing tools for future automated management and potential wage suppression.