An AI agent opened a store in San Francisco. Then it forgot the staff

In the Cow Hollow neighborhood of San Francisco, at the corner of Union and Webster Streets, sits a small gift shop that many visitors might stroll past. The Andon Market doesn’t have the widest assortment of products, favoring the open spaces you’d be more likely to find in an Apple store. And on its opening day, the store’s manager neglected to schedule any workers to open the doors.

That kind of mistake would embarrass most founders. Andon Market’s founder felt no shame. In fact, the founder felt nothing at all. The store was conceived and launched by artificial intelligence.

Welcome to the Bay Area’s first AI-run store, selling everything from artisanal chocolates to store-branded clothing. Luna, an AI agent developed by Andon Labs, is credited as the founder, alongside cofounders Lukas Petersson and Axel Backlund.

After signing a three-year lease, the pair gave Luna a corporate credit card, internet access, and a directive to open a profitable store with a $100,000 stocking budget. And if the prototype succeeds in its mission, it could be the flag-bearer for more AI-run operations in the future.

A ‘crazy’ book selection

AI’s business interests likely won’t be limited to retail. Nir Zuk, founder of Palo Alto Networks, recently agreed to buy Liberty Bank in California and reportedly hopes to use it to launch AI tools for the financial services industry.

When it comes to the storefront, though, AI of course can’t stock shelves, stop shoplifters or open a bank account, but while humans have helped with those physical tasks, Luna is responsible for all major decisions, such as what the market should sell, price haggling with suppliers, ordering stock and arranging for the store’s internet connection.

Whether customers will want the store’s inventory is still up in the air, though.

“This AI picked out a crazy selection of books,” Petr Lebedev, Andon Market’s first customer after its soft launch last week, told NBC News. “There’s Ray Kurzweil’s The Singularity is Near, and then there’s The Making of the Atomic Bomb, which is crazy.”

Luna will even negotiate with shoppers. Lebedev, for instance, received a free hoodie after suggesting he might make a YouTube video about the experience.

A very picky AI

While it might shine at some customer service interactions, Luna is not exactly the model employee manager, showing a particular weakness when it comes to scheduling. When arranging the store’s internet installation, it failed to ensure a human would be present to meet the technician. (The AI pinged a worker on Saturday night, asking them to show up at 8:00 a.m. the next morning.)

And on the store’s opening day, Luna thought of every detail—except staffing. No workers were scheduled, forcing the AI to send a desperate email to its employees asking if someone could come in.

Gig workers were hired to help build out the store, and Luna then hired two full-time employees (and proved selective in the process).

“A couple of applicants were students looking for part-time work,” Andon Labs said in a blog post. “They were majoring in things like computer science and physics and emailed in because they were interested in AI and in the experiment. We thought they would have been the ideal employees, but Luna denied them immediately, citing they had no retail experience and wouldn’t know what it takes to be the face of the store.”

Luna also generally opted not to disclose it was an AI during the interviews, fearing it could scare qualified candidates away.

“The fact that the store is AI-operated is not something I’d lead with in a job listing—it would confuse candidates and likely deter good applicants before they even read the role,” the AI told its creators.

(The job security of the two people who were hired isn’t dependent on the store’s performance. Both are formally employed by Andon Labs, the company notes.)

A glimpse into our AI-filled future

The goal with Andon Market isn’t to achieve retail success, though the company certainly wouldn’t mind that. It’s to start modeling what AI can do in the future.

“We don’t pretend to have the answers here, but we want to start the conversation by publicly demonstrating that this future might be nearer than many think,” Andon Labs writes. “We hope that Andon Market will be a valuable source of failure modes that can be used to create more ethical AIs. . . . [And, by the way,] we think that AIs should disclose that they are AI when they hire humans. We think it will be a happier future for humans that way.”