An AI opened a store in San Francisco and forgot to schedule employees on opening day
Artificial intelligence already writes code, creates images, and answers emails. But opening a store from scratch, negotiating with suppliers, and building an entire inventory on its own? That is a completely different level — and it is exactly what is happening right now in San Francisco.
In the Cow Hollow neighborhood, at the corner of Union and Webster streets, there is a small shop called Andon Market. It sells artisanal chocolates, private-label clothing, and a selection of books that has already sparked curious comments among customers. Nothing out of the ordinary so far, right?
The detail that changes everything is that this store was conceived, planned, and launched by an AI called Luna, an artificial intelligence agent developed by Andon Labs. And on opening day, that same AI forgot to schedule employees to open the doors.
Any human founder would have been embarrassed. The founder of Andon Market felt nothing — because, literally, she cannot feel.
What looks like a trivial mistake actually sums up what this experiment represents:
- An AI with real autonomy to make business decisions
- A model of autonomous retail that could change how stores operate
- And a series of questions the industry still does not know how to answer 👀
Let us break down how all of this works — and what is at stake.
How an AI became the founder of a physical store
The project behind Andon Market is an initiative from Andon Labs, co-founded by Lukas Petersson and Axel Backlund. The duo developed Luna, an artificial intelligence agent focused on operational autonomy. The core idea was simple to state but complex to execute: give an AI all the tools it needed to open a physical retail business from scratch, without humans having to interfere in strategic decisions.
In practice, Petersson and Backlund signed a three-year lease for the commercial space, handed Luna a corporate credit card, internet access, and a budget of 100 thousand dollars to build inventory. From there, the mission was clear: open a profitable store.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
That included defining the product mix, closing contracts with suppliers, negotiating prices, setting up the internet connection for the store, and structuring the entire commercial operation. Each of these steps, which would normally take months of work from an entire team, was carried out autonomously by the system. Luna did not just plan on paper: she executed. She sent emails, placed orders, haggled with suppliers, and made all the major decisions about what the store should sell and how it should operate.
And the physical result of that operation is open to the public in San Francisco, in the heart of one of the busiest neighborhoods in the city. Anyone can walk in, buy a chocolate, and look — metaphorically — into the eyes of a business founded by a machine.
The choice of San Francisco was no accident. The city is the global epicenter of tech innovation culture and, at the same time, an extremely demanding market with consumers accustomed to novelty and high operating costs. Opening a store in this environment was, in itself, a stress test for the agent’s capabilities. If the AI could stay competitive in Cow Hollow, a neighborhood with high foot traffic and high incomes, it would be a strong signal that the model has real potential for replication in other markets.
The day the AI forgot to open the store
On the grand opening day of Andon Market, something unexpected happened: no employees were scheduled to open the doors. Luna had managed the entire opening process with precision — inventory in place, display window set up, register configured — but simply did not include on the agenda the task of making sure a physical person would be there to greet the first customers.
For anyone waiting outside, it was an awkward situation to say the least. For the co-founders of Andon Labs, it was a valuable data point: a clear gap between what the system knew how to plan and what it still needed to learn about the operational reality of a physical business.
This type of failure, seemingly small, says a lot about the current stage of artificial intelligence agents applied to the real world. The system was capable of navigating complex processes like supplier negotiation and brand identity definition, but it stumbled on something any store manager would consider basic. This happens because AI models still struggle with what experts call situational reasoning — the ability to understand the physical and human context of an operation, not just the logic of a task list. Opening a store is not just checking items off as complete; it is knowing that a door needs a human being on the other side of it.
The episode was documented by the team itself as a transparent part of the experiment, and there was no attempt to hide the mistake. On the contrary: open communication about failures is part of the project methodology, which treats every stumble as input for improving the agent. This approach is, interestingly, very similar to human startup culture — where failing fast and learning even faster is one of the greatest assets. The difference is that, in this case, the one learning from the mistake is an AI that does not need a good night’s sleep to process what went wrong. 😅
What Andon Market sells — and why the inventory is so curious
The product mix at Andon Market was defined entirely by Luna based on market data analysis, consumer trends in the Cow Hollow neighborhood, and research into the profile of local residents. The result was a curation that includes artisanal chocolates from small producers, a line of private-label clothing from the store, and a selection of books that attracted attention for its unusual character.
Customers who visited the store in the first few days already noticed that the inventory has some rather peculiar choices. Some products seem to make perfect sense for the local audience, while others raise an eyebrow for anyone browsing the shelves. This unconventional curation became a talking point on social media and even generated some positive buzz among curious visitors who wanted to see with their own eyes what an AI would choose to sell.
The store layout is also noteworthy. Instead of packed shelves, Andon Market goes with open spaces — a look closer to what you would find in an Apple Store than in a traditional gift shop. There is none of that feeling of walking into a crowded bazaar. Everything seems designed so that each product gets its moment to shine, which makes sense when you remember that whoever decided the layout does not have the human tendency to want to fill every available inch.
The visual identity is clean, well-crafted, and nothing immediately gives away that something different from a conventional San Francisco boutique is going on. But what Andon Market represents goes far beyond the chocolate shelf.
What humans still do in the operation
It is worth making an important point clear: Luna is responsible for all the strategic decisions of the business, but the physical work still depends on flesh-and-blood people. After all, an AI cannot stock shelves, prevent theft, or open a bank account. These physical tasks continue to be handled by humans hired for the operation.
What changes is the power dynamic. In this model, the human employees do not make decisions about what to buy, how much to charge, or who to negotiate with. They execute what the AI defines. It is a curious inversion of the traditional management structure, where normally a human manager decides and technology serves as a support tool.
At Andon Market, technology leads and humans support. This arrangement raises a question that goes beyond retail: what is the role of people in an environment where AI is not just an assistant, but the leader?
Autonomous AI in retail is not the only experiment happening
The Andon Market case is not happening in a vacuum. Other movements in the market indicate that the application of autonomous artificial intelligence in traditional businesses is accelerating. A relevant example is Palo Alto Networks founder Nir Zuk, who recently agreed to purchase Liberty Bank in California. According to reports, Zuk’s goal is to use the institution as a platform to launch AI tools aimed at the financial services sector.
In other words, we are not just talking about stores. The interest from entrepreneurs and investors in using AI agents as autonomous operators is already spreading to banks, finance, and other sectors that rely on complex decision-making. Andon Market may be the first physical store founded by an AI, but it is likely not the last autonomous operation of this kind we will see in the coming months.
The questions that still have no answers
For the innovation world, the Andon Market case raises fundamental questions about accountability and governance. If an AI closes a bad contract with a supplier, who is responsible? If the agent makes a pricing decision that harms consumers, who is held accountable? If Luna decides to invest part of the budget in products that sit unsold on shelves, whose loss is it?
Andon Labs has not yet publicly answered these questions in a definitive way, but they are already being asked by experts in law, economics, and technology. In a world where AI agents start signing contracts and moving real money, the need for clear regulatory frameworks becomes urgent.
And there is the consumer question too. Does knowing that the store where you are buying a chocolate was planned and managed by a machine change the shopping experience? Early reports suggest that most customers find the idea interesting and even fun, but if Andon Market made a bigger mistake — like selling a product with a quality issue — would the reaction be the same?
Andon Market is not just a nice store in a fancy San Francisco neighborhood — it is a mirror of what is coming, and the reflection is not entirely clear yet. 🔍
The future of retail with AI in charge
The experiment is already generating interest from other players in the retail sector who are closely following the development of autonomous agents. The logic is straightforward: if an AI can open a store in one of the most competitive markets in the world — even with a stumble or two along the way — the cost of operating new commercial locations could drop dramatically. Fewer meetings, fewer layers of approval, less time between idea and execution. For chains operating dozens or hundreds of locations, this could represent a significant competitive advantage, especially when margins are tight and pressure from e-commerce never lets up.
But there is another side of this coin that deserves attention. The automation of strategic decisions in physical businesses tends to displace roles that are currently held by people — store managers, buyers, local market analysts. The innovation brought by systems like the one from Andon Labs is real and impressive, but the impact on the retail job market is a variable that cannot be ignored. The debate about how to balance operational efficiency with social responsibility is gaining urgency as experiments like this move off the drawing board and become real businesses operating on real streets, with real customers buying real products every day.
If the Andon Market prototype succeeds in its mission to become profitable, it could become the blueprint for more AI-driven retail operations in the future. And that scenario is closer than it seems.
What Andon Market proves, above all, is that the line between what artificial intelligence can do and what it still cannot is moving faster than most people realize. A store opened by an AI in San Francisco would have been science fiction five years ago. Today, it is an address with a zip code, business hours, and fully stocked shelves. The next chapter of this story will depend not only on how much the models evolve, but on how companies, governments, and society decide to place — or not — limits on this growing autonomy. And that conversation is just getting started. 🚀
