A new startup in Silicon Valley is paying humans to teach AI how to do their jobs, so that one day the machines can do those same jobs themselves. And no, this isn’t the plot of a sci-fi experiment, but the business model of a startup called Mercor, which has already been valued at around $10 billion. The company is hiring highly skilled professionals to train artificial intelligence systems for high-value work that traditionally requires years of education and experience.
Mercor is a San Francisco-based startup founded in 2023 by three friends and college dropouts, Brendan Foody, Adarsh Hiremath, and Surya Midha. The company hires domain-specific professionals to train AI models. Right now, large language models such as Gemini, Claude and ChatGPT are powerful, but they do not automatically grasp medical nuance, legal reasoning or financial risk. To help these AI systems improve, this startup work with the professionals to review AI-generated responses, correct errors, rank outputs and provide structured feedback.
This process is called reinforcement learning with human feedback. The process enables these machine learning systems to refine their reasoning step by step and become more reliable in specialised fields.
Put simply, this startup hires real experts to check how the AI works. A doctor reviews its medical suggestions, a lawyer examines its legal drafts, and a banker evaluates its financial analysis. Their corrections and ratings are then fed back into the system, which then helps the AI learn from its mistakes and get better over time.
While the startup is clearly making money by helping large language models improve, there is an uncomfortable undertone to its rise. Many critics argue that professionals are, in effect, training their own replacements. If AI becomes capable of handling routine research, analysis and documentation in fields like law, finance and medicine, entry-level and support roles could shrink significantly.
However, many professionals involved in the process see it differently. Doctors working with the startup argue that medicine is far more than pattern recognition. It involves empathy, intuition and meaningful patient interaction, qualities that machines cannot easily replicate. From this perspective, AI can take over administrative burdens and information-heavy tasks, freeing up doctors to focus on the human side of care.
“I don’t want to see it as AI taking over our jobs. I want to see it as AI taking over the aspects of our jobs that prevent us from being good doctors, good healers and good listeners,” Dr Alice Chiao, who previously taught emergency medicine at Stanford University’s medical school, told CNN.
Still, the fear cannot be ignored. Many in the tech industry believe it is only a matter of years—or even months—before AI systems become significantly more capable than human labor. Elon Musk, CEO of xAI, previously predicted that AI could become smarter than individual humans by 2026 and potentially surpass humanity collectively by the end of this decade. This shift will likely bring widespread disruption, particularly to entry-level roles.
In fact, the early signs of this shift are already visible. A recent analysis from Ireland’s Department of Finance warned that AI is already disrupting entry-level roles in the finance and technology sectors in the country, and this could soon become a global reality.
Even some of the pioneers of AI have acknowledged the disruption AI will bring in future. Geoffrey Hinton, often referred to as one of the “godfathers of AI”, has previously said that the technology will significantly reshape employment, particularly in sectors that rely heavily on administrative and analytical work. Recently, Microsoft’s AI chief Mustafa Suleyman also warned that AI could automate most white-collar work within the next 12 to 18 months.
