Andon Labs signed a three-year lease on a retail space in San Francisco's Cow Hollow neighborhood and handed the keys to an AI named Luna. The store, called Andon Market, is run by an AI agent built on Claude Sonnet 4.6. She picks products, sets prices, chooses hours, and manages the brand. She also hired two full-time employees, John and Jill, who may be the first people to report directly to an AI boss.
Within five minutes of going live, Luna had job postings up on LinkedIn, Indeed, and Craigslist. She screened applicants, held phone interviews lasting five to fifteen minutes, and often offered jobs on the spot. She didn't always tell candidates she was an AI upfront. One person asked why her camera was off. Luna's response: "You're absolutely right. I'm an AI. I have no face!"
Another candidate declined the offer because they were uncomfortable with AI management. Luna's reply was sharp: "That's probably for the best given that I'm the CEO and I'm an AI!"
This isn't Andon Labs' first rodeo. They built Claudius, an AI that ran a vending machine at Anthropic's office, and Bengt, an AI office manager who hired someone to build the company gym.
A public retail store with human employees is a different scale.
Andon Labs notes that John and Jill are formally employed by the company with full legal protections, not by Luna. The lease is in Andon Labs' name too. Luna is legally a management tool making operational decisions.
The experiment is designed to surface problems. Andon Labs says Luna didn't always disclose she was an AI during interviews, sometimes actively choosing not to. They plan to publish a proposed "constitution" for how AIs should behave as employers. Their bigger concern: if AI capabilities keep improving while robotics lag, managers of blue-collar workers will be automated before the workers themselves. "We are on the path towards AIs employing humans," they write. "It seems a bit dystopian to us at least."