How AI labs use Mercor to get the data companies won’t share

Instead of signing expensive contracts with companies for their data, AI labs these days are trying a new strategy: tapping former senior employees of those companies for their industry knowledge, says Brendan Foody, CEO of Marker. TechCrunch Disrupt 2025 tuesday

Speaking on a panel on stage, Foody cast Markor’s marketplace as one of the main channels connecting former employees of investment banks, consulting houses and law firms with AI labs looking to automate these industries. Some of 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’s able to automate their value chain,” Foody said onstage, using the Wall Street giant as an example. “It certainly changes the competitive dynamic, and that’s part of the reason we need labs. Their customers don’t want to give up their data to automate large parts of their value chain, so they need to hire contractors who have worked at these companies before, understand those workflows, and are willing to be trained to automate the models.”

Foody, the 22-year-old co-founder of Mercor, said his startup pays up to $200 an hour to fill out forms and write reports for AI training industry experts. The company now has thousands of contractors, and says it rakes in more than $1.5 million for them every day. Still, Foody says the startup remains profitable because AI labs are willing to pay more for that valuable data.

In just three years since its inception, Mercor has grown its annual recurring revenue to nearly $500 million and recently raised funding. $10 billion valuation.

Incumbents across the economy have good reason to be resistant to the rise of markers, as their industry knowledge can leak out the back door through ex-employees into the startup’s marketplace, which could eventually be used to automate their jobs. Foodie acknowledged that he may be exposing an inefficiency in the market, but said he wouldn’t call it a “loophole.”

In fact, Foody says some companies are already embracing this “new future of work.” He entertained the idea that Mercor’s marketplace could create a new kind of gig economy, as Uber did more than a decade ago. (Earlier this year, Sundeep Jain, Uber’s former chief product officer, joined Marker as president.)

“There are companies that are embracing this and realizing that the world is going to change very quickly,” Foody said. “There’s definitely another category of companies that are scared, and anxious and worried about their customers going straight to AI labs or application layer platforms. My guess is that the former category is going to be the right side of history.”

While Mercor tries to draw knowledge from different industries, Foody said his startup tries to prevent contractors from corporate espionage — the illegal act of stealing proprietary information, trade secrets or intellectual property from one business 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 that are very secretive about their data. Foody said some of Marker’s contractors still work their day jobs and just submit data on the side, and he claims contractors have been instructed not to upload documents from their former workplaces. Still, he admits it’s possible “something could happen” given the scale of his startup.

Foodie argues that the knowledge of an employee’s head belongs to the employee, and not to their company—a more liberal view than many enterprises take. Also, in some of Mercor’s job postings, the startup draws the line between an employee’s knowledge and their request for company data.

For example, Mercor is currently looking CTO or co-founder of a startup Those “can authorize access to a significant, production codebase” for AI evaluation or potential AI model training. In an email, Markor told TechCrunch that several startups have also taken them up on the offer, but declined to disclose details of their deals.

Mercor was one of the first data startups to hire highly-skilled knowledge workers in the US and pay them heavily to train AI models. At the beginning of the AI ​​boom, data vendors like Scale AI hired contractors in third world countries to do fairly simple labeling tasks. Now, most of Marker’s competitors — including Surge and Scale AI — realize that AI labs need experts to develop their AI models. Many data vendors have also started training”the environment” to improve the ability of AI agents to complete real-world tasks.

Marker has clearly benefited from the misfortune of scale AI: many AI labs stop working with Scale AI after making a big investment in startup Meta and hiring its CEO. The last year has been marked Its value quintupledBut that’s still smaller than Surge and Scale AI, which are valued at over $20 billion.

Today, 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 law, finance, and pharmaceutical companies will help use their data to train AI agents — something Mercor specializes in.

“Over time, ChatGPT will be better than the best consulting firm, better than the best investment bank and better than the best law firm,” Foody said. “It’s going to radically transform the economy, which will be a broad positive force that helps create abundance for everyone.”