Shift Offers Free NYC Apartment Cleanings in Exchange for Robot Training Data — 10,000 Operators, 15 Countries
MicroAGI has a straightforward pitch: your apartment gets cleaned for free, and the footage of cleaners scrubbing, mopping, and organising goes toward training the robots that will eventually replace the cleaners.
The company, operating under the brand Shift, launched the offer in New York City on May 29. The mechanism is a hat-mounted camera — the company calls it a “magic hat” — worn by the cleaner during the job. It captures a first-person point-of-view record of every task: bathroom scrubbing, floor mopping, laundry folding, dish washing, refrigerator organisation. The footage is anonymised before use: faces, names, and personal information visible on screens or ID cards are blurred.
That footage is then licensed to AI and robotics companies as training data for embodied AI systems.
Why first-person matters
Egocentric video is the highest-signal format for training robotic manipulation systems. Third-person footage captures what happened; first-person footage captures how — hand positions, camera motion, the micro-adjustments humans make when a surface is slippery or an object is oddly shaped. That data feeds behaviour cloning and imitation-learning pipelines in ways that simulation cannot replicate at scale.
Shift’s pitch to robotics labs is that real-world residential data — messy, variable, uncontrolled — is more valuable for generalisation than clean simulation. The website notes: “more challenging cleaning environments can be especially useful.”
The business model
MicroAGI already operates a broader data-collection platform. More than 10,000 operators across 15 countries are paid to record everyday household and professional tasks through the Shift app, earning approximately $20 per hour plus bonuses. The company paid out more than $5 million in Q1 2026.
The free cleaning service in NYC is both a data acquisition strategy and a recruiting play. Dozens of blog posts targeting NYC students, restaurant workers, and neighborhood residents appear on the company’s site. The unit economics reportedly work: the training data value per cleaning session exceeds the cost of paying the gig worker.
Expansion to San Francisco, London, Zurich, and Munich is described as “very soon.”
The catch
The cleaners are not Shift employees — they are vetted gig workers through third-party partners, and the terms of service disclaim liability for property damage, theft, or personal injury. A payment method is required at booking; cancellations with less than 24 hours’ notice incur a charge.
The free cleaning offer is described as “limited time.” The primary purpose of the Shift app is ongoing: recruiting operators to record daily activities for pay, with the NYC free-cleaning campaign serving as a high-visibility funnel.
Context
Shift is one of several companies attempting to industrialise embodied AI training data collection. Others in the space include Encord and Micro1, which has hired contract workers across 50 countries. What distinguishes Shift is the consumer-facing barter model: instead of recruiting contributors directly, it turns households into data collection sites by subsidising a service those households already pay for.
The robots that training data is meant to enable do not yet exist at commercial scale. But the labs building them need millions of hours of manipulation footage before they will. That data scarcity is, right now, worth more per apartment than a professional cleaning costs.