How AI Labs uses Mercor to capture data that companies won’t share

AI labs are trying something new these days: hiring former senior employees at companies to gain industry knowledge instead of signing expensive contracts with companies for data, Mercor CEO Brendan Foody said Tuesday at TechCrunch Disrupt 2025.

Speaking on an on-stage panel, Foody cited Mercor’s marketplace as one of the key channels to connect former employees of investment banks, consulting firms, and law firms with AI labs looking to automate those industries. Mercor’s customers include OpenAI, Anthropic, and Meta.

“There’s an argument that Goldman Sachs doesn’t like the idea of ​​having a model that can automate the value chain,” Foudy said, citing Wall Street giants as an example. “Competitive dynamics are definitely changing, and that’s part of why labs need us. Customers don’t want to provide data to automate large parts of their value chain, so they need to hire contractors who have previously worked for those companies and are willing to understand those workflows and train models to automate them.”

Mercor co-founder Foody, 22, says his startup pays industry experts up to $200 an hour to fill out forms and write reports for AI training. The company currently employs tens of thousands of contractors and pays them more than $1.5 million every day. Still, Foudy says the startup remains profitable because the AI ​​lab is willing to pay more for its valuable data.

In less than three years since its founding, Mercor has grown annual recurring revenue to approximately $500 million and recently raised funding at a $10 billion valuation.

There are good reasons for economic incumbents to resist Melkor’s rise. Industry knowledge could be leaked out the back door through former Melkor marketplace employees and ultimately used to automate tasks. Foudy acknowledged that it may be exposing market inefficiencies, but said he wouldn’t call it a “loophole.”

In fact, some companies are already beginning to embrace this “new future of work,” Foody says. He was intrigued by the idea that Mercor’s marketplace could create a new type of gig economy, much like Uber did more than a decade ago. (Earlier this year, Sundeep Jain, Uber’s former chief product officer, joined Mercor as president.)

“Some companies are embracing this and recognizing that the world is changing rapidly,” Foudy said. “There is definitely another category of companies that are afraid and concerned about disintermediation being cut out and customers having direct access to AI labs and application layer platforms. My hunch is that the former category will be on the right side of history.”

While Melkor seeks to draw knowledge from a variety of industries, Foudy said his startup is trying to stop corporate espionage by contractors, the illegal practice of stealing confidential information, trade secrets or intellectual property from one company and selling it to another.

But that’s easier said than done. Most of Mercor’s employees are former employees of law firms, investment banks, and other industries where data is highly sensitive. Foudy said some Melkor contractors still work full-time jobs and only submit data on the side, and claimed that contractors have been instructed not to upload documents from their previous workplaces. Still, he acknowledged that given the size of his startup, “something could happen.”

Foody argues that the knowledge in an employee’s head belongs to the employee, not the company. This is a more generous view than many companies take. Additionally, in some of Mercor’s job postings, the startup walks the line between requesting employee knowledge and company data.

For example, Mercor is currently looking for a CTO or co-founder of a startup “who can authorize access to a substantial production codebase” for AI evaluation or potentially training AI models. Melkor told TechCrunch in an email that several startup CTOs have taken up the offer, but declined to reveal the details of the deal.

Mercor was one of the first data startups in the US to hire highly skilled knowledge workers and pay them handsomely to train their AI models. In the early days of the AI ​​boom, data vendors like Scale AI hired contractors from third-world countries to do fairly simple labeling tasks. Now, most of Mercor’s competitors (including Surge and Scale AI) have realized that their AI labs need experts to improve their AI models. Many data vendors have also begun training “environments” to improve the ability of their AI agents to complete real-world tasks.

Mercor is clearly benefiting from Scale AI’s misfortune. After Meta invested heavily in the startup and hired a CEO, many AI labs stopped working with Scale AI. In the last year, Mercor has increased its value five times, but it remains smaller than Surge and Scale AI, both valued at more than $20 billion.

Currently, most of Mercor’s revenue comes from just a few AI labs, but Foody says the startup plans to partner with other industries in the future. He believes legal, financial, and medical companies need help leveraging data to train the AI ​​agents that Mercor specializes in.

“Over time, ChatGPT will be better than the best consulting firm, better than the best investment bank, better than the best law firm,” Foody said. “It will be a far-reaching positive force that will fundamentally transform the economy and create prosperity for all.”