The proposition sounds almost too convenient: A professional will clean your New York apartment for free. They'll tackle the dishes piled in the sink, vacuum under the couch, maybe even organize that pantry you've been avoiding since February. No charge.
The tradeoff? A camera strapped to their chest records every second of it.
That's the pitch from MicroAGI, a German startup that launched its Shift service in Manhattan on May 29, 2026. The business model is bracingly simple: Record first-person video of humans doing household chores, then license that footage to companies building the next wave of home robots. In exchange for letting cameras into your kitchen—and bedroom, and bathroom—you get a service that normally costs $150 or more, completely free.
According to Semafor, which first covered the launch, Shift drew "thousands and thousands of bookings" in its opening days. Apparently New Yorkers, a population not known for volunteering their privacy lightly, were willing to make the trade.
What MicroAGI is really selling, of course, isn't cleaning. It's chaos. The unscripted clutter of actual homes—the dog bowls on the floor, the toddler toys scattered across the living room, the pile of mail on the counter—represents exactly the kind of messy, real-world data that robotics labs struggle to simulate at scale. For AI developers training systems to navigate physical space and manipulate objects, footage of human hands wiping down a sticky countertop or folding a fitted sheet is worth subsidizing someone else's housework.
"We record first-person cleaning footage… That data is valuable enough for us to offer cleaning services free of charge for a limited time," the company explains on its consumer site. Harry Kilberg, Shift's U.S. general manager, framed it to Semafor as a way to "democratize the AI economy," allowing someone to clean a neighbor's apartment while Shift pays for the recording rights.
Whether "democratize" is the right word for a system that might eventually automate away the very jobs it's currently recording is a question the company hasn't dwelled on publicly.
The Marketplace Mechanics
Shift positions itself as a marketplace connecting consumers with cleaning professionals through its booking site, shiftapp.nyc, though details about its operational structure remain sparse. The company vets cleaners via partners (details remain sparse), equips them with cameras, motion sensors, and GPS trackers, then dispatches them into homes. Clients specify tasks, consent to filming, and understand the preferential pricing hinges entirely on that consent.
The data haul extends well beyond video. According to Shift's privacy policy, posted June 2, 2026, the company also collects inertial measurement unit (IMU) data, GPS coordinates, camera metadata, phone specifications, and the cleaner's height. All of it gets bundled into datasets, processed on cloud servers in Belgium, and licensed to what MicroAGI describes as AI and robotics customers. Specific client names? The company hasn't disclosed any.
MicroAGI advertises on its operator recruitment portal, joinshift.us, that it paid out more than $5 million to contributors in the first quarter of 2026 alone, though this figure hasn't been independently verified. The startup also advertises operations across 15-plus countries and partnerships with over 10,000 businesses and households—claims that similarly lack external validation. Dealroom lists MicroAGI, based in Munich, with 11 to 50 employees and no known external funding as of its last update—making the company's reported Q1 payout claim all the more striking, if accurate.
Why Robots Need Your Mess
The appetite for this data reflects a particular challenge in robotics development. Training systems to fold laundry, wipe counters, or maneuver through a kitchen crowded with pets and yesterday's takeout containers requires enormous volumes of first-person footage from uncontrolled environments. Labs can stage scenarios, but there's no substitute for the unpredictability of real homes.
Shift isn't pioneering this approach so much as scaling it aggressively. Build AI released a 100,000-hour egocentric video dataset in late 2025. Mecka AI, also based in New York, raised $60 million in June 2026 to capture human motion data through body sensors and iPhones—footage intended for robotics training. Over in Shenzhen, X Square Robot teamed up with 58.com earlier this year to deploy human-robot cleaning crews in actual apartments, blending service delivery with data harvesting.
A Fortune newsletter covering Mecka's funding round singled out MicroAGI's free-cleaning gambit as a standout example of the embodied-data trend. Digital Trends put it more bluntly: "Your messy home is valuable AI training data."
For developers training vision-language-action models or imitation learning systems, the footage of real hands manipulating real objects in real spaces is worth paying for. Or, in Shift's calculus, worth giving away a cleaning service to get.
Privacy by Design—With Caveats
Shift's privacy framework leans on automated blurring of faces, heads, license plates, screens, IDs, and payment cards to produce what it calls "reduced-identifiability" datasets. The company's privacy policy, labeled Draft v0.1 and last updated June 2, 2026, is refreshingly candid about the limitations: "We do not represent that recordings are fully or permanently anonymous." Home interiors, after all, can be identifying even without faces.
Consent is required before recording begins. Clients are responsible for notifying other adults in the household; cleaners are instructed to avoid filming minors or intimate situations. Either party can pause or stop recording mid-session. Clients can request deletion before footage is incorporated into external datasets, but once the data ships to AI labs, removal becomes murky. The policy acknowledges the company will delete recordings "to the extent technically possible" after distribution.
According to the privacy policy, data is encrypted in transit and at rest, stored and processed in the EU under SHIFT APP US LLC, a Delaware entity running the New York program, though these claims have not been independently verified. The company says it doesn't sell personal information for advertising, but it does license datasets to its own subsidiaries and third-party developers.
The Cool Down, a climate and tech outlet, reported the launch was "met with major skepticism," referencing the OECD.AI database's classification of the program as an "AI Hazard" citing privacy risk, though the specific documentation of this classification is not detailed. Commentary across tech blogs has raised a parallel concern: the data being collected could help build the very automation that displaces the low-wage cleaners currently wearing the cameras.
Perhaps ironically, the humans doing the recording may be training their own replacements.
What Comes Next
Kilberg told Semafor that Shift has plans to expand beyond cleaning into plumbing, cooking, and construction work, and into additional cities, though these remain aspirational rather than confirmed implementations. The current New York program is billed as a limited-time offer, though no sunset date has been announced. Legal terms posted June 3 suggest future pricing models could blend free, discounted, and paid tiers depending on whether customers consent to recording.
For now, the New York pilot is a test—of consumer tolerance, regulatory ambiguity, and the economics of trading privacy for convenience. MicroAGI's wager is that the value of a free professional cleaning outweighs the discomfort of cameras in the bedroom, or the knowledge that your home layout might end up in a dataset licensed to a dozen robotics labs.
Whether that calculus holds as the service expands, and as more people grasp what "reduced-identifiability" actually means in practice, is an open question. But the early booking surge suggests the company has struck a nerve. Or maybe just a bargain too good to refuse.
New Yorkers, it turns out, will let the robots watch them work—so long as someone else does the scrubbing.
