Mercor pays professionals to help train AI on their own jobs

In short: Mercor is building a large paid network of professionals who help train AI systems by reviewing and improving the AI’s work in their own fields.

What's going on

Mercor is a San Francisco startup that pays white-collar professionals to help train AI systems on job-related tasks. This includes work tied to law, medicine, finance, writing, and other areas where experience matters.

Multiple reports describe Mercor as a $10 billion company with about 30,000 contractors. Reported pay varies by role, but coverage and company materials cite high rates, often in the $75 to $200+ per hour range, with averages commonly reported around $85 to $105 per hour.

Mercor says the work is not about replacing a lawyer or doctor directly. Instead, contractors review, rate, and improve AI outputs using detailed guidelines. Think of it like being a paid editor and grader for an AI, where your job is to show what a good answer looks like and explain why.

Journalists and the company describe this as a new kind of gig economy for knowledge work. In this setup, experts lend their judgment to AI labs such as OpenAI and Anthropic through a middle platform.

What to watch

The big question is what this work turns into over time. One possibility is that AI handles routine tasks, while humans focus on tricky cases and supervision. Another is that as AI improves, companies may hire fewer deep experts and more lower-cost general reviewers. A third possibility is that if AI becomes reliable enough, the need for human correction could shrink, which is why some see these high-paying gigs as both a short-term opportunity and a long-term warning.

Source: NYTimes