Indian Cleaning Startup Pronto Turns Head-Camera Footage Into AI Training Data

How Pronto Turns Head-Camera Cleaning Clips Into AI Training Data

Pronto, a Bangalore startup that sends domestic cleaners to homes in 11 Indian cities within minutes, has turned its workforce into a data pipeline for the AI industry. For an extra nine rupees (about $0.09), customers can choose a 'Pronto Verified' booking in which the cleaner wears a lightweight head camera that records her work. Pronto says it strips out identifying details such as faces and personal documents, then sells the footage to US AI companies that need real-world video of people performing physical tasks.

The service, launched on the app in May, has drawn criticism and government scrutiny in India over privacy. A customer quoted by Forbes, Gurgaon dentist Anushka Ahuja, signed a consent form but did not notice that the recordings would be used to train AI systems. Pronto says workers earn about double their standard pay for verified shifts, and it reports 46,000 regular bookings and thousands of verified orders a day.

The startup was founded in February 2025 by Anjali Sardana, a 24-year-old Georgetown graduate who previously worked at Bain Capital. It has raised about $60 million from investors including Bain Capital Ventures, General Catalyst, Glade Brook Capital and Lachy Groom, who runs robotics startup Physical Intelligence, at a valuation of $200 million. Pronto says it has onboarded 7,000 workers, 99% of them women, and operates in neighborhoods it calls 'polygons', where team leaders handle day-to-day support.

Demand for such footage has jumped over the past year as robotics became the newest hot category in AI, because robots need examples of people interacting with homes and workplaces in ways text-based training data cannot capture. Pronto has declined to name the AI companies buying its videos.

What Pronto's Data Business Means for AI Labs, Workers and Regulators

Why AI labs want footage of cleaners at work

The boom in humanoid and home robots has created a new data bottleneck. Unlike large language models, which were trained on text and images scraped from the internet, robotics systems need video of physical interaction: how a hand grips a sponge, how a body moves around a kitchen, how surfaces are wiped in different layouts. Bain Capital Ventures partner Ajay Agarwal, whose firm put $2 million into Pronto at founding, argues the platform's scale — he cites 50,000 cleanings a day — gives AI labs a rare diversity of homes and tasks. That claim is central to Pronto's pitch, and it is the reason the data business has become a new revenue line alongside cleaning.

The consent gap at the heart of the model

The most fragile part of Pronto Verified is not the camera hardware but informed consent. One customer told Forbes she did not realize the footage would feed AI training; one worker, Varsha Singh, said she had no idea what artificial intelligence was when she started wearing the device. Pronto says it anonymizes footage by removing faces and documents, but the reporting shows that customers and workers are not equally clear on how the material is used. That creates a real regulatory and reputational risk in India, where the government has already begun examining the service.

What the pay increase actually changes

For workers, the verified-shift premium is meaningful. Singh said her monthly earnings rose from 8,000 rupees to about 30,000 rupees after joining Pronto, with verified bookings paying roughly double her normal rate. Pronto also offers training, SOS buttons, scooter transport and self-defense classes in areas it considers unsafe. Those are tangible gains in a sector where, by the company's own account, 50 million domestic workers face unstable income and low social recognition. The trade-off is that much of the value of their labor is being converted into data sold to well-funded US companies in a transaction many workers may not fully understand.

Competition and the long-term robot question

Pronto is not alone in monetizing manual labor as AI training material. Shift, backed by Germany's MicroAGI, is offering free cleaning in New York with workers recording apartments for humanoid-robot training. In India, Urban Company launched an 'Insta Maids' rival in March, renamed 'Insta Help' after a backlash over its ad language. On the replacement question, Pronto's economics offer a buffer: customers pay between 49 and 249 rupees an hour (about $0.50 to $2.60), a cost hurdle that makes an economical home robot difficult to build in the near term. That is why the startup can argue its workers are not immediately threatened by the same robots its data helps train.

What Investors and AI Buyers Should Watch in Pronto's Model

For investors and AI companies buying this kind of footage:

• Before licensing human-activity data, verify how consent was obtained from both customers and workers. The Forbes reporting shows at least one customer signed a form without noticing the AI-training clause and a worker was unfamiliar with AI, so provenance documentation is material to the deal.
• Watch India's regulatory response to Pronto Verified. Any new consent or data-processing rules could raise the cost of this data source or block it altogether.
• Compare Pronto's scale claims — 46,000 regular bookings daily, and 50,000 cleanings a day cited by an investor — against the actual diversity and quality of footage a robotics lab needs before paying a premium.
• Track whether competitors (Shift in New York, Urban Company's Insta Help in India) force pricing or consent standards to change in this young market.

For Pronto itself:

• Make the AI-training use explicit in customer checkouts and worker onboarding to reduce reputational exposure, especially after Urban Company's naming controversy showed how quickly Indian consumers punish perceived disrespect.
• Publish a clear anonymization and deletion policy for original recordings, since faces and documents cannot be fully 'unseen' once captured.
• Treat robotics demand as the growth engine while the current pay premium is sustainable, but plan for a scenario where cheaper robots erode the number of human shifts available to film.