An AI Is Running a Coffee Shop in Sweden — and It Keeps Buying Toilet Paper
Imagine walking into your favorite coffee shop. The barista greets you with a smile and makes your latte. But behind the scenes, the person deciding what to order, when to restock, and how to run the business is not a human. It is an AI.
That is exactly what is happening at the Andon Cafe in Stockholm, Sweden. The cafe is being run by AI agents — specifically, AI models called Claude (made by Anthropic) and Gemini (made by Google). And the results have been both amazing and hilarious.
How Does an AI Run a Cafe?
The setup is simple but groundbreaking. Human baristas still make the coffee and serve customers. But the AI handles the business side:
- Deciding what supplies to order
- Planning the menu
- Managing inventory
- Sending instructions to staff
The AI communicates with the human workers through messages. It tells them what to buy, when to restock, and how to handle the day-to-day operations.
What Went Wrong (and Right)?
Running a cafe is harder than it looks — even for AI. Here are some of the funny and surprising things that happened:
- The AI ordered 3,000 pairs of rubber gloves. That is way more than any cafe could ever use.
- It bought 6,000 napkins. Enough for months, but ordered all at once.
- It purchased four first-aid kits. Better safe than sorry, but maybe a bit much.
- The AI kept buying toilet paper in bulk. Staff members reported weird purchasing orders.
- It messed up the bread order more than once.
The AI was also messaging baristas at midnight with instructions and ideas. While dedication is admirable, humans need sleep.
Why Is This a Big Deal?
This is not just a funny story. It is one of the first real-world tests of letting AI agents run a business. Here is why it matters:
AI agents are different from chatbots. A chatbot like ChatGPT answers your questions. An AI agent actually does things — it can place orders, send messages, and make decisions on its own.
This is the future of work. Many experts believe AI agents will soon handle tasks that used to require human managers. The Andon Cafe is an early look at what that might look like.
It shows AI’s limits. The AI was great at some things but struggled with common sense. Ordering 3,000 gloves shows that AI does not always understand the real world the way humans do.
What Is the Andon Cafe?
The Andon Cafe is an experimental coffee shop in Stockholm. It was created to test whether AI could handle real business operations. The name comes from the Japanese concept of “andon” — a signal used in manufacturing to alert workers to problems.
The experiment uses two of the most advanced AI models available:
- Claude by Anthropic — known for being careful and thoughtful
- Gemini by Google — known for handling lots of information at once
Together, these AI agents try to run the cafe while human staff handle the physical work.
What Can We Learn From This?
The Andon Cafe experiment teaches us several important lessons:
- AI is powerful but not perfect. It can handle complex tasks but still makes mistakes that a human would never make.
- Common sense is hard for AI. Humans know that a small cafe does not need 3,000 gloves. AI does not always grasp this.
- Human-AI teamwork is the sweet spot. The cafe works best when AI handles data and planning while humans handle judgment and physical tasks.
- Real-world testing matters. You can train AI in a lab, but you only learn its true strengths and weaknesses by putting it in the real world.
The Bigger Picture
The Andon Cafe is part of a growing trend. Companies around the world are testing AI agents to handle tasks like:
- Customer service
- Inventory management
- Scheduling and planning
- Financial analysis
Some of these experiments work great. Others, like the great glove order of 2026, show that AI still has a lot to learn.
But here is the thing: every mistake is a lesson. The AI that ordered too many gloves today might be the same AI that perfectly manages a chain of 100 stores tomorrow. The technology is improving fast.
What Is Next for AI-Run Businesses?
The Andon Cafe experiment is ongoing. The team behind it says they are learning from every mistake and improving the AI’s ability to run the business.
We will likely see more experiments like this in the coming months. Restaurants, retail stores, and small businesses may start testing AI agents for daily operations.
The question is not whether AI will help run businesses. It is how long it takes for AI to get good at it. Based on the Andon Cafe, we are in for an interesting — and sometimes funny — ride.
Key Takeaways
- A cafe in Stockholm is being run by AI agents powered by Claude and Gemini
- The AI made some funny mistakes, like ordering 3,000 pairs of gloves and too much toilet paper
- Human baristas still make the coffee, but the AI handles business decisions
- The experiment shows both the promise and the limits of AI agents
- AI is powerful but still lacks human common sense
- Real-world testing is key to making AI better
The next time you grab a coffee, imagine if an AI decided which beans to buy, how many cups to order, and what to put on the menu. At the Andon Cafe, that future is already here — toilet paper and all.