All Knowledge Work Is Becoming AI Training, Mercor’s CEO Explains the Inevitable Shift

Chen Ling
2 days ago
6 min read

Artificial intelligence is no longer just a tool that boosts productivity or automates routine tasks. In 2026, it is actively redefining what work means, who performs it, and how value is assigned to human knowledge. From the perspective of Mercor’s CEO Brendan Foody, this transformation is not about machines replacing people, but about the economy reorganizing itself around a new core activity, teaching machines how to think.

Mercor, a three-year-old company that has rapidly grown into a central player in AI training infrastructure, sits at the intersection of elite human expertise and frontier AI development. Foody’s view offers a rare inside look at how labor markets are being reshaped from the ground up, not by layoffs, but by a fundamental shift in what counts as valuable work.

This article rewrites and expands the earlier analysis by placing Mercor’s CEO perspective at the center, explaining why AI labs increasingly depend on top-tier professionals, why only a small fraction of human contributors drive most AI progress, and why Foody believes all knowledge work is converging toward training AI agents.

From Traditional Careers to AI Knowledge Markets

For decades, professional success followed a familiar arc. Individuals acquired education, joined prestigious institutions, accumulated experience, and delivered outputs such as reports, strategies, legal opinions, or financial models. Value was measured by what those outputs achieved for clients or employers.

Foody argues that AI disrupts this model at a structural level. In his view, the most valuable output of expert work is no longer the document or decision itself, but the reasoning process behind it. That reasoning, when captured and transferred to an AI system, becomes reusable at scale.

From Mercor’s perspective, this is why AI labs no longer rely primarily on crowdsourced labor. Large language models and autonomous agents do not fail because they lack data, they fail because they lack judgment. Judgment comes from people who have spent years making high-stakes decisions under uncertainty.

Mercor was built around this insight. Instead of optimizing for volume, the company optimizes for expertise density, connecting AI labs with professionals who have operated at the highest levels of finance, consulting, law, engineering, and policy.

Why Elite Expertise Outperforms Massive Scale

One of Foody’s most striking observations is that AI improvement does not scale linearly with the number of contributors. According to his experience, a small minority of contributors account for the majority of meaningful model gains.

From Mercor’s internal data and client outcomes, several patterns consistently emerge:

The top 10 to 20 percent of expert contributors drive most reasoning improvements
High-quality feedback reduces model errors faster than large volumes of generic data
Expert corrections generalize across tasks, while low-skill inputs remain narrow

Foody often emphasizes that this mirrors real-world organizations. In most companies, a small group of top performers disproportionately shapes outcomes. AI training, he argues, is no different.

This is why Mercor invests heavily in screening and evaluation. The goal is not to find people who can answer questions, but people who can explain why an answer is correct, what assumptions it relies on, and when it might fail. These meta-cognitive skills are what AI systems struggle to learn on their own.

The Economic Logic Behind Paying Experts to Train AI

At first glance, it seems paradoxical that professionals are paid to train systems that may later automate parts of their own industry. Foody does not see this as contradictory. From his point of view, it is simply rational market behavior.

Expert knowledge has a limited window of maximum value. As AI systems improve, routine applications of that knowledge become cheaper. The peak value moment is when expertise is still scarce and AI systems urgently need it.

Mercor’s model allows professionals to monetize their expertise at that peak. Instead of competing with automation later, they arbitrage their knowledge earlier, converting years of experience into direct income.

Foody frames this as a shift from labor as execution to labor as instruction. Historically, professionals were paid to perform tasks repeatedly. In the AI era, they are increasingly paid to explain how tasks should be performed, and why.

The Boundary Between Personal Expertise and Corporate Knowledge

A recurring concern around Mercor’s model is whether former employees are effectively transferring corporate intellectual property into AI systems. Foody addresses this directly and acknowledges the complexity.

From his perspective, there is a clear distinction between proprietary secrets and generalized professional judgment. Mercor does not allow contributors to share confidential documents, client data, or internal processes tied to specific firms. What contributors provide instead is abstracted reasoning, patterns, and frameworks that exist independently of any one employer.

Foody argues that most competitive advantage in knowledge industries has never come from secrecy alone. It comes from execution, culture, and continuous innovation. AI accelerates the diffusion of generalized expertise, but it does not eliminate the need for institutions to adapt and evolve.

This creates discomfort for traditional firms, but Foody sees it as inevitable. Once knowledge exists in human minds, it cannot be permanently contained. AI simply makes that diffusion more visible and more scalable.

Why Crowdsourcing Breaks Down at the Frontier

Earlier generations of AI benefited from crowdsourced labor because the tasks were well-defined and easily verifiable. Labeling images, transcribing speech, or ranking search results could be distributed to large groups with minimal context.

Foody explains that frontier AI systems fail in different ways. They hallucinate, reason incorrectly, or apply correct logic in the wrong context. Fixing these problems requires deep domain understanding.

From Mercor’s CEO viewpoint, crowdsourcing breaks down because:

Contributors lack the context to diagnose errors
Feedback becomes superficial rather than explanatory
Models learn surface patterns instead of principles

Expert contributors, by contrast, can explain why a model’s answer is wrong, what assumption failed, and how to reason correctly under uncertainty. This type of input compounds in value as models scale.

An AI system trained with expert reasoning does not just perform better on one task, it becomes more robust across entire classes of problems.

Foody’s Core Belief, All Knowledge Work Converges on AI Training

Perhaps the most provocative element of Foody’s worldview is his belief that all knowledge work eventually converges on training AI agents. He does not mean that everyone becomes a data labeler. He means that the primary economic output of knowledge professionals shifts from human-facing deliverables to machine-facing instruction.

In this future:

Analysts focus on edge cases and failure modes
Lawyers teach systems how to reason under conflicting precedents
Consultants encode strategic trade-offs rather than slide decks
Engineers supervise systems that design other systems

Foody sees this as a long transition, not an overnight replacement. For years, humans and AI will coexist in hybrid workflows. But the direction is clear, the center of gravity moves from doing to teaching.

This view reframes AI anxiety. Instead of asking which jobs disappear, Foody asks which parts of jobs become instructional, supervisory, or ethical in nature.

Implications for Workers, Power, and Inequality

From Mercor’s CEO perspective, AI-driven labor markets will reward excellence more than ever. This creates opportunity, but also risk.

On one hand, elite professionals gain unprecedented leverage. They can work across industries, geographies, and organizations, monetizing their expertise directly. On the other hand, mid-level and routine knowledge work faces increasing pressure as AI absorbs standardized tasks.

Foody does not claim this outcome is inherently fair. He acknowledges that AI amplifies existing inequalities by concentrating value among top performers. However, he also argues that this concentration already existed, AI merely exposes it.

The policy challenge, in his view, is not to stop AI training markets, but to expand access to high-level skill development so more people can participate meaningfully.

How Institutions Must Adapt

For companies, Mercor’s rise signals a structural change. Knowledge is no longer confined within organizational walls. Expertise is fluid, portable, and increasingly monetized externally.

Foody believes institutions that thrive will:

Treat learning as a continuous process
Focus on proprietary data and execution speed
Integrate AI training into core workflows
Redefine loyalty around purpose, not exclusivity

Those that cling to rigid hierarchies and knowledge hoarding will struggle as AI accelerates talent mobility.

Work After AI, Through Mercor’s Lens

From Brendan Foody’s perspective, AI is not ending work, it is clarifying it. The value of human labor is shifting away from repetitive execution toward judgment, explanation, and ethical guidance. Mercor exists because AI systems need the best of human thinking before they can operate autonomously.

In 2026, the most important question is no longer which jobs AI will replace, but whose knowledge will shape the machines that replace them. The answer determines who gets paid, who holds influence, and who defines the future of intelligent systems.

As AI continues to reshape global labor markets, insights from leaders like Foody help illuminate where value is truly moving. For readers seeking deeper strategic analysis on AI, labor, and global technological power shifts, further perspectives from Dr. Shahid Masood and the expert research team at 1950.ai provide critical context on how these transformations affect economies, governance, and long-term human agency.