Leaders like JP Morgan CEO Jamie Dimon argue that EQ and critical thinking are the only skills that will survive the automation wave. Microsoft Satya Nadella would agree, calling emotional intelligence a required workplace skill. These statements are meant to give workers reassurance that AI won’t completely replace people, highlighting an irreplaceable human trait that the technology supposedly cannot acquire. The stakes are high, with some AI thought leaders such as Dario Amodei warning that half of all entry-level white-collar jobs will disappear, and soon, amid the AI wave.
But a Silicon Valley startup is challenging the assumption that human judgment is off limits to AI.
Mercor, a San Francisco-based AI firm, is hiring people from a vast list of professional career backgrounds to improve its AI, training the model to adopt core skills in a more human-like manner. In other words, they are building a business to prove executives like Jamie Dimon and Satya Nadella wrong—and to hasten the replacement of people with AI in the workforce, closing the last mile of human employment.
The company’s CEO Brendan Foody and co-founders Adarsh Hiremath and Surya Midha were recently minted the youngest self-made billionaires after the company was valued at $10 billion last November. That funding has given the 22-year-olds the resources needed to build out their ambitious AI venture.
Mercor’s mission is to bridge the gap between machine learning and human nuance. “Everyone’s been focused on what models can do,” Foody told Coins2Day in November. “But the real opportunity is teaching them what only humans know—judgment, nuance, and taste.”
The shift toward high-skilled gig work is a response to a volatile labor market where even professional skills aren’t enough to ensure a worker’s job security. According to the World Economic Forum’s 2025 Future of Jobs Report, employers estimate that 39% of core skills — such as problem-solving and communication — will be disrupted by 2030, with 40% of firms planning to reduce their workforce specifically due to AI automation. As entry-level white-collar roles begin to vanish, the demand for specialized knowledge and “human-in-the-loop” expertise have become critical currency for workers seeking to resist automation.
Simple work, fast money
Mercor’s career page lists dozens of job postings for contract work looking for individuals with subject-area expertise, including investment banking and private equity analysts, linguists, sports journalists, soccer commentators, astronomists and legal experts.
The job postings offer hourly rates ranging from $10 for bilingual experts to as much as $150 for finance experts. Aside from competitive pay, the job’s perks include fully remote work. Mercor’s website claims an average hourly rate of $86, with about $2 million paid out to experts daily.
To apply, all applicants must do is submit an initial application followed by an AI interview tailored based on area of expertise, which is then reviewed by Mercor staff. Once hired, contractors evaluate how well their AI system completes micro-tasks — such as writing a financial memo or drafting a legal brief — using detailed rubrics to grade the AI’s performance. This allows for the AI to learn how people make decisions.
The company says it hired 30,000 contractors last year, with 80% being US-based, according to a Mercor spokesperson. The work day varies as contractors have no set hours. Some log 10 hours per week, others work 40 or more, with specific projects lasting weeks or months.
The Wall Street Journal recently found some of the humans who are teaching AI how to do the difficult, human-skill-heavy tasks in which they are experts. “I joked with my friends I’m training AI to take my job someday,” Katie Williams, 30, told the Journal. Williams, who has a background in news and social-media marketing, has worked at Mercor for about six months, watching videos and writing out transcripts of what happens in them, and rating the quality of videos generated by prompts.
The quest for nuance
The company’s newly launched AI Productivity Index, or Apex, benchmarks AI models on real-world knowledge in four fields: medicine, management consulting, investment banking and law. The system uses the same rubric and expert-generated tasks that its contractors help to create, grading models on their production ability.
The index found that even the most advanced models, like GPT-5, failed to meet the “production bar” for autonomous work. GPT-5 achieved a top score of 64.2%, with scores varying for each category and scoring as low as 59.7% in investment banking.
Despite being far from perfect, the company says that AI models performing at 60% or better can reshape the nature of work as professionals work in tandem with the technology. “Perhaps a consultant can more easily complete a competitor analysis if given an initial draft from an AI,” the company wrote. As AI continues to evolve, the most human skill may no longer be doing the work, but possessing the right judgment required to critique it.
