Micro AGI sends free cleaning staff to NYC apartments for AI training

There’s an old saying in tech: if the product is free, you’re the product. MicroAGI, a Munich-based startup, is putting a very literal spin on that idea. The company launched its Shift app on May 28, offering free professional apartment cleanings across New York City. The catch: every cleaner wears a head-mounted camera, recording first-person footage of themselves scrubbing your countertops, mopping your floors, and folding your laundry.

That footage, once anonymized to strip out identifiable information, becomes training data for the next generation of household robots.

How the data pipeline works

The core problem MicroAGI is trying to solve is actually straightforward. Teaching a robot to navigate a real apartment and perform physical tasks requires enormous quantities of real-world visual data. Simulated environments only get you so far. At some point, the AI needs footage of actual human hands doing actual human chores in actual cluttered apartments.

That’s where Shift comes in. New Yorkers sign up through the app, a professional cleaner shows up for free, and the entire session is recorded through a head-mounted camera. MicroAGI says the footage is then anonymized, meaning faces, personal items, and other identifying details are removed before the data enters their training pipeline.

The company then sells these anonymized datasets to AI and robotics firms that need high-quality physical task data. It also uses the footage internally for its own research and development on what it calls “end-to-end physical AGI.”

A global operation disguised as a cleaning service

The free NYC cleanings grabbed headlines, but MicroAGI’s data collection operation is far bigger than one city. The company has built a contributor network of more than 10,000 operators spread across 15 countries. These operators are compensated roughly $20 per hour to record themselves performing everyday tasks while wearing the company’s camera rigs.

This network generated more than $5 million in revenue during the first quarter of 2026 alone. MicroAGI was established in 2025 and is headquartered in Munich, with operations spanning Istanbul, Zurich, London, and New York. The company also offers research fellowships that provide up to $2 million in compute resources per fellow.

Why this matters for the AI and robotics market

The broader context here is a well-known bottleneck in the robotics industry. Language models like GPT had the entire internet’s worth of text to train on. Image generators had billions of photos and illustrations scraped from the web. But physical AI, the kind that needs to understand how to grasp a dish towel or navigate around a coffee table, doesn’t have an equivalent data reservoir.

MicroAGI’s approach flips the script by turning everyday service work into a scalable data pipeline. Instead of building expensive lab environments, the company sends workers into real homes, capturing first-person, unscripted footage that makes training data genuinely useful.

There are obvious risks, though. Privacy is the elephant in the room. Even with anonymization, recording the inside of someone’s apartment generates sensitive data. Regulatory scrutiny seems inevitable as the operation scales, particularly in the European Union, where GDPR enforcement has teeth.

There’s also the question of whether the “free cleaning” model is sustainable or just a marketing play to generate buzz during the NYC launch. Paying 10,000 operators $20 per hour globally is a significant labor cost, and the unit economics only work if the datasets command premium prices from buyers.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.