Startup offers free home cleaning in NYC to gather AI training data for future household robots

MicroAGI Launches Free Home Cleaning Program to Collect Robot Training Data

German startup MicroAGI is offering New York City residents free home cleaning services through its newly launched Shift app, but there's a significant trade-off involved. The company sends professional cleaners wearing camera-equipped headsets to record first-person footage of every task they perform, from scrubbing dishes to mopping floors 1

1

. This AI training data will be used to teach the next generation of AI-driven robots how to navigate domestic chores 3

3

.

The unusual pitch, announced on May 28 with a promotional video set to Jay-Z and Alicia Keys' "Empire State of Mind," positions the free cleaning offer as a mutually beneficial arrangement. According to the Shift app website, the value of the training data collected is substantial enough to justify covering the cost of professional cleaners 1

1

. MicroAGI founder and CEO Bercan Kilic has teased plans to expand the service to additional cities including London, Munich, and Zurich 4

4

.

Why Physical World Data Creates Bottlenecks for Household Robots

Unlike chatbots and image generators that exploded in popularity by scraping text and images from the internet, robots must contend with the physical world. This requires understanding space, motion, force, friction, unusual shapes and materials, awkward lighting, and countless other variables that humans grasp instuitively 2

2

. Tasks that seem simple for people—folding clothes, picking up an apple, pouring water—have proven extraordinarily difficult for roboticists to codify.

Data collection for robotics presents a massive bottleneck for companies developing physical AI. The physical world is harder to scrape than digital content, and harder still to scrape quietly without compensating people who generate it 2

2

. This scarcity makes egocentric footage particularly valuable, as it captures exactly how humans navigate physical space from a first-person perspective.

The Shift app already claims to pay tens of thousands of people across 15 countries to record everyday household or professional tasks, with the main platform suggesting more than 10,000 operators have collectively earned over $5 million in the first quarter of the 2026 fiscal year 1

1

. Contributors reportedly earn $20 per hour plus bonuses for wearing recording headstraps to capture short videos of daily activities 1

1

.

Privacy Concerns and Data Anonymization Promises

The Shift app's privacy policy states that advanced machine learning models running directly on smart glasses or video capture devices perform irreversible transformations such as automated face blurring and identifier obfuscation before any data uploads to cloud servers 1

1

. The company promises to blur all personally identifiable information from screens, ID cards, pieces of paper, and cell phones to protect both customers and their homes 5

5

.

However, significant questions remain unanswered. There's no mention of whether people can request removal of their home cleaning videos from training datasets for robots after recording 4

4

. It's also unclear whether the company's data anonymization techniques adequately prevent homes from being identified when footage appears in training datasets 1

1

. Even with good intentions, data breaches happen regularly, raising concerns about videos of private homes potentially spreading across the internet 5

5

.

The Shift app terms of service also attempt to absolve the platform of responsibility for property damage, theft, or personal injury that may occur during cleaning appointments, though the company states cleaners are vetted by its partners 1

1

.

Growing Market for Robotics Training Through Human Activity Recording

MicroAGI joins a growing number of startups recruiting ordinary people to record everyday tasks for robot training purposes. Other companies pursuing similar strategies include Encord and Micro1, with the latter having hired thousands of contract workers across 50 countries including India, Nigeria, and Argentina 1

1

.

In India, home services platform Pronto has been using clients' homes as a source of AI training footage for domestic chores like cooking, cleaning, and laundry, though the practice triggered backlash with rival startups insisting they never recorded inside homes 2

2

. Silicon Valley-based Human Archive partners with companies to have gig workers record activities using camera caps that collect first-person perspective footage 2

2

.

Some companies have established staged data farms where workers are paid to complete identical physical tasks repeatedly while cameras and sensors capture every movement, turning rote activities like folding towels and picking up cups into valuable training material 2

2

.

What This Means for the Future of Domestic Labor

The free home cleaning services represent only a limited-time promotional offer, though booking requires payment information and customers may face charges for cancellations with less than 24 hours' notice 1

1

. MicroAGI has launched an aggressive recruiting campaign with dozens of blog posts targeting NYC university students, teachers, restaurant workers, delivery workers, and residents of specific neighborhoods, while spreading Craigslist postings to other US cities like Boston 1

1

.

Cleaning may serve as just the beginning. The company's promotional video indicates plans to eventually expand into other areas including plumbing, cooking, and building 3

3

. The company's FAQ notes that more challenging cleaning environments prove especially useful for training future robots 3

3

.

This approach to gathering physical world data highlights how companies are willing to offer free services in exchange for access to information that could eventually enable them to sell automated solutions back to consumers. The act of trading data for value isn't new—loyalty cards, cookies, dashcams, and insurance apps have long monitored behavior in exchange for discounts or convenience. What's shifted is the type of data companies now consider worth paying for, and the creative methods they're deploying to obtain it 2

2

.