AI companies have already scraped much of the web for training data. Now they want data that shows how humans actually work, according to a Business Insider series called The AI Data Grab.
Some pay thousands of workers worldwide to record everyday tasks so robots can learn physical jobs. Others build virtual workplaces where AI agents practice workflows, make decisions and recover from errors.
Contract workers in Mexico and the Philippines remotely control robots in American warehouses as they pack beauty products, recording movements so the machines can eventually do the work on their own. Workers elsewhere strap cameras to their foreheads to capture first-person footage of folding laundry and making hotel beds, Business Insider reported on Oct. 15.
The physical data rush
“This is how the great knowledge transfer happens from humans to robots,” said Josh Gruenstein, CEO of Tutor Intelligence. His company operates a Massachusetts “data factory” where workers remotely control robots to generate training data.
Demand for robot training data is roughly 10 times higher than a year ago, said Vineeth Velmurugan, who leads robot learning at the AI data startup Encord. Encord’s customers, including Dyna Robotics and Agility Robotics, once requested 1,000 hours of annotated data. Now they want 10,000 or even 100,000, he said.
Microagi, the German parent company of Shift, is expanding its collection efforts into homes and workplaces across more than 15 countries. It pays more than 35,000 people to record themselves performing physical tasks, CEO Bercan Kilic said. Shift said participants can earn up to $30 an hour and that it has paid out around $10 million in 2026.
Digital workplaces for AI agents
The push extends to software agents. Tech companies are building realistic digital workplaces, called reinforcement learning environments or RL environments, where AI can practice coding, using business software, making decisions and completing long-running tasks much like a human employee.
Google was in talks to invest more than $1.5 billion in Mechanize, a startup that builds virtual work environments to train AI agents, Business Insider reported in August. Meta has been trying to collect employees’ keystrokes, mouse movements, clicks and other screen activity, saying the goal is to help AI learn how people actually use computers.
At Scale AI, nearly half of new AI training projects now involve RL environments that model coding, computer use and enterprise workflows, Chetan Rane, its head of product for agents and RL environments, wrote in a blog post. Instead of judging a single answer, these environments let an AI attempt an entire workflow, make mistakes, recover and improve based on whether it completed the task.
Company records are part of the same hunt. AI firms are offering businesses six- and even seven-figure sums for anonymized Slack messages, emails, presentations and internal documents, to teach AI how people solve problems, collaborate and handle long, multi-step tasks, according to another story in the series. Law firms, meanwhile, are building AI systems around decades of their own legal knowledge rather than relying solely on off-the-shelf models.
In an Oct. 18 overview of the series, Business Insider’s Alistair Barr wrote that the first wave of generative AI was mostly about bigger models and faster chips. “The next chapter is about machines understanding how work gets done, so AI can help us do it better and quicker,” he wrote.