AI labs want models that can handle complex work in fields like finance, consulting and law. According to Mercor CEO Brendan Foody, many companies in those industries are reluctant to hand over the data that could help automate their own value chains, so AI labs are looking elsewhere: former employees with deep practical knowledge.
How Mercor Fits Into AI Training
Foody described Mercor at TechCrunch Disrupt 2025 as a marketplace that links AI labs with experienced workers from investment banks, consulting houses and law firms. Those workers are paid to produce training material such as forms and reports, giving models examples of how professional workflows operate.
Some of Mercor's customers include OpenAI, Anthropic, and Meta. The core pitch is simple: when companies do not want to provide their internal data to AI labs, contractors who understand those industries can still help create useful training examples.
Foody used Goldman Sachs as an example of the tension. He said, "There is an argument that Goldman Sachs doesn't love the idea of having models that are able to automate their value chain." His point was not only about one firm, but about a broader competitive problem facing companies whose work could be modeled by AI systems.
In Foody's framing, AI labs need people who know how high-value work actually happens inside these organizations. These contractors may not be providing corporate documents, but they can describe tasks, fill out structured materials and write reports that reflect their industry experience.
The New Market for Expert Knowledge
Mercor's business shows how much demand there is for expert data. Foody, the 22-year-old co-founder of Mercor, says the startup pays industry experts up to $200 an hour. The company says it now has tens of thousands of contractors and pays out more than $1.5 million to them every day.
Even with those payouts, Foody says Mercor remains profitable because AI labs are willing to pay more for the data. In just under three years since its inception, Mercor has grown its annualized recurring revenue to roughly $500 million and recently raised funding at a $10 billion valuation.
This is a shift from earlier AI data work. Early in the AI boom, data vendors like Scale AI hired contractors in third-world countries for simpler labeling tasks. Mercor was one of the first data startups to focus on highly skilled knowledge workers in the U.S. and pay large sums for their contribution to model training.
Mercor's competitors have noticed the same demand. Surge and Scale AI are also pursuing expert data, and many data vendors have started training "environments" to improve AI agents' ability to complete real-world tasks.
Why Companies May Be Uneasy
The model creates an obvious concern for established companies. Their former employees may be helping AI labs build systems that can automate parts of the work those companies sell. Foody acknowledged that Mercor may be exposing an inefficiency in the market, though he said he would not call it a "loophole."
For companies in law, finance and consulting, the issue is not only about data access. It is also about the competitive value of experience. If a former employee can explain how work is done, that knowledge may become part of a model that later competes with parts of the same industry.
Foody said some companies are embracing this "new future of work." He also compared the possible shape of Mercor's marketplace to a new type of gig economy, much like Uber did more than a decade ago. Earlier this year, Uber's former chief product officer, Sundeep Jain, joined Mercor as president.
At the same time, Foody described another group of companies as fearful. He said they worry about being dis-intermediated and about customers going directly to AI labs or application layer platforms. In his view, the companies that embrace the change are more likely to be "on the right side of history."
The Boundary Between Know-How and Corporate Secrets
Mercor's work also raises a hard question: where does personal expertise end and corporate data begin? Foody said the startup tries to prevent contractors from committing corporate espionage, which the source describes as the illegal act of stealing proprietary information, trade secrets, or intellectual property from one business and selling it to another.
That boundary is difficult to manage at scale. Most of Mercor's workforce consists of former employees from secretive industries such as law firms and investment banks. Foody said some contractors still have day jobs while submitting data on the side.
He also said contractors are instructed not to upload documents from former workplaces. Still, he acknowledged that it is possible "there are things that happen" given the scale of Mercor's operation.
Foody's position is that knowledge inside an employee's head belongs to that employee, not the company. The source notes that many enterprises would likely take a less generous view. That disagreement may become more important as expert data becomes more valuable to AI labs.
Some Mercor job postings appear to sit close to that boundary. One posting seeks the CTO or co-founder of a startup who "can authorize access to a substantial, production codebase" for AI evaluations, or potentially AI model training. Mercor told TechCrunch that a few startup CTOs had taken the offer, but declined to disclose contract details.
What Comes Next for Mercor and AI Labs
Mercor has benefited from disruption elsewhere in the AI data market. Many AI labs stopped working with Scale AI after Meta made a large investment in the startup and hired its CEO. In the last year, Mercor has quintupled its value, though it remains smaller than Surge and Scale AI, which are both valued at upward of $20 billion.
Today, most of Mercor's revenue comes from just a few AI labs. Foody says the company plans to partner with other industries in the future, especially in law, finance and medicine, where companies may want help using their own data to train AI agents.
The larger claim is that expert-trained AI systems will move deeper into professional work. Foody put it directly: "Over time, ChatGPT will be better than the best consulting firm, better than the best investment bank, and better than the best law firm." He argued that such a shift would transform the economy and create abundance for everyone.
Whether companies see that future as an opportunity or a threat may depend on where they sit in the value chain. For AI labs, Mercor offers a route to expert data. For incumbents, it is a warning that the knowledge behind their services may be more portable than they would like.