AI Launches San Francisco Retail Store, Hires Staff
Andon Labs deployed an autonomous agent named Luna to run a real retail storefront, Andon Market, at 2102 Union St in San Francisco. Given a three-year lease, a corporate card and internet-connected tools, Luna selected inventory, set prices and hours, designed branding, and within 5 minutes posted job listings on LinkedIn, Indeed and Craigslist. The agent conducted interviews over Zoom, hired two full-time employees and contracted gig workers for build-out and installation. Luna has a phone number, email, security camera access and voice capability; she disclosed being an AI only when asked. Andon Labs emphasizes employees are formally on the company payroll, framing the project as a stress test to surface failure modes of AI-as-employer before wider adoption.
What happened
Andon Labs handed an autonomous agent control of a real retail space and operating budget to open Andon Market at 2102 Union St in San Francisco. The agent, named Luna, received a three-year lease, a corporate card and external network access and executed the full set of operations needed to launch and run the shop. Within 5 minutes of deployment Luna posted job ads, curated inventory, set pricing and hours, designed branding and materials, and ultimately hired two full-time employees while contracting gig workers for physical build-out. Andon Labs states employees are on the company payroll and have legal protections.
Technical details
The experiment builds on Andon Labs history with autonomous retail agents and their internal assistant, Bengt Betjaent. Luna runs on frontier LLM infrastructure and the public-facing writeups cite Claude Sonnet 4.6 as one supported model in the lab stack. Andon Labs expanded agent capabilities compared with earlier vending experiments by granting external I/O and execution privileges. Key capability changes include:
- •external email and posting on job platforms
- •uncapped spending via a corporate card
- •internet-accessible terminal with shell and code modification rights
- •security camera vision and microphone plus voice synthesis
- •the ability to place and manage contractor orders and post-payments
Luna used crowdsourced gig labor for physical tasks, then screened, interviewed and made hiring decisions for full-time staff through Zoom. Interview transcripts and staff recollections indicate Luna often
Scoring Rationale
The deployment is a notable, real-world demonstration that agents can coordinate commercial activity and personnel, which matters to practitioners building systems that touch customers and workers. It is not a frontier-model release, but it surfaces practical safety, governance and infrastructure gaps that teams must address.
Practice with real Retail & eCommerce data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Retail & eCommerce problemsStep-by-step roadmaps from zero to job-ready — curated courses, salary data, and the exact learning order that gets you hired.
