AI Reshaping Work: The Startling Transformation of Who Gets Paid for Knowledge in 2026

BitcoinWorld

AI Reshaping Work: The Startling Transformation of Who Gets Paid for Knowledge in 2026

In a startling shift for the global labor market, artificial intelligence is not just automating tasks but fundamentally redefining who performs valuable work and how they are compensated. According to Brendan Foody, CEO of the three-year-old startup Mercor, we are witnessing the emergence of a new economic layer where elite human expertise fuels the very AI systems poised to transform entire industries. This conversation, captured at the 2026 Disrupt conference, reveals how a $10 billion company has positioned itself at the epicenter of this transformation by connecting top-tier professionals with leading AI labs.

AI Reshaping Work Through a New Kind of Middleman

Mercor operates as a specialized intermediary in what CEO Brendan Foody describes as AI’s ongoing data gold rush. The company’s model is straightforward yet revolutionary. Instead of relying on traditional crowdsourced labor platforms, Mercor specifically recruits former employees from prestigious institutions like Goldman Sachs, McKinsey & Company, and elite law firms. These individuals then provide their deep industry expertise to train advanced AI models developed by labs such as OpenAI and Anthropic.

Consequently, these experts command significant fees, sometimes reaching $200 per hour, for sharing knowledge that could ultimately lead to the automation of roles within their former sectors. This creates a complex economic paradox where professionals are paid to train their potential replacements. The rise of this model coincides with broader industry shifts, including the well-publicized operational challenges faced by competitor Scale AI, which inadvertently accelerated Mercor’s ascent by highlighting the limitations of conventional data-labeling approaches for complex, knowledge-intensive tasks.

The Critical Role of High-Skilled Contractors

Foody’s central thesis challenges the assumption that more data always leads to better AI. He argues that the quality of training data, derived from human expertise, matters far more than quantity. In his analysis, the top 10-20% of contractors on platforms like Mercor drive the majority of measurable model improvement. These individuals possess not just rote knowledge but nuanced understanding, critical judgment, and the ability to contextualize information within real-world scenarios.

Therefore, Mercor’s entire operational focus revolves around identifying and vetting this elite cohort. The company employs sophisticated screening processes that go beyond resumes, assessing problem-solving approaches and domain-specific insight. This focus on quality over scale represents a significant departure from earlier waves of AI training, which often utilized vast pools of lower-cost, generalist labor. The economic implication is clear: a premium is now placed on specialized human intelligence as the key feedstock for artificial intelligence.

Navigating the Gray Area of Corporate Knowledge

A critical and contentious aspect of this new economy involves the boundary between an employee’s personal expertise and a former employer’s proprietary secrets. When a former investment banker trains an AI on financial modeling, where does general professional knowledge end and corporate intellectual property begin? Foody acknowledges this gray area, stating that Mercor has strict protocols and legal frameworks to prevent the sharing of confidential information.

However, the existential question for firms like Goldman Sachs remains. Should they be worried that the collective expertise of their alumni is being used to build AI agents capable of performing high-finance tasks? Industry analysts suggest that while outright theft of trade secrets is illegal, the aggregation and synthesis of publicly available knowledge and generalized professional skills by AI poses a longer-term, strategic disruption threat to all knowledge-based industries.

The Inevitable Convergence: All Knowledge Work as AI Training

Looking forward, Foody presents a provocative vision: he believes the entire knowledge economy will eventually converge on the activity of training AI agents. In this future state, the primary output of many professional roles will not be a report, a legal brief, or a financial model for human consumption, but rather curated data used to refine and improve autonomous AI systems. This represents a fundamental reorientation of work’s purpose.

This transition follows a logical, if disruptive, trajectory. Initially, AI needs human experts to learn foundational concepts and complex reasoning. Subsequently, as models become more capable, the human role evolves from direct task execution to oversight, correction, and teaching—focusing on edge cases, ethical considerations, and strategic direction that the AI cannot yet grasp independently. This timeline suggests a prolonged period where human expertise becomes increasingly valuable as a supervisory and pedagogical resource, even as AI handles more routine analytical functions.

Conclusion

The insights from Mercor’s CEO reveal that AI reshaping work is less about simple job displacement and more about the radical reconfiguration of value creation. A new market has emerged that monetizes elite human cognition directly as a training input for machine intelligence. This shift challenges traditional corporate structures, intellectual property norms, and career pathways. As Foody predicts, if all knowledge work ultimately becomes training data, then the most critical skill for the future professional may not be domain mastery alone, but the ability to effectively transfer that mastery to the next generation of non-human intelligence.

FAQs

Q1: What exactly does Mercor do?
Mercor is a platform that connects highly skilled professionals, often from top-tier firms in finance, consulting, and law, with artificial intelligence research labs. These professionals are paid to share their expert knowledge to train and improve advanced AI models.

Q2: Why do AI labs need high-skilled contractors instead of cheaper, crowdsourced labor?
According to Mercor’s CEO, the quality of training data from top experts leads to significantly better AI model performance. The nuanced judgment and deep contextual understanding of elite professionals drive more substantial improvements than larger volumes of input from generalist crowdsourced workers.

Q3: Is it legal for former employees to train AI with knowledge from their old jobs?
There is a legal gray area. While sharing specific trade secrets or confidential client data is illegal, general professional skills and publicly available industry knowledge are not proprietary. Mercor states it has protocols to prevent the sharing of confidential information.

Q4: What is the long-term vision that Mercor’s CEO describes?
Brendan Foody believes that eventually, all knowledge-based work will transform into the activity of training AI agents. The primary output of many professional roles will shift from direct work products to curated data used to teach and refine autonomous AI systems.

Q5: How did Scale AI’s troubles affect Mercor?
Reports of operational and quality control challenges at Scale AI, a major data-labeling platform, highlighted the limitations of traditional crowdsourcing for complex AI training needs. This contrast helped accelerate interest in and adoption of Mercor’s quality-focused, expert-driven model.

This post AI Reshaping Work: The Startling Transformation of Who Gets Paid for Knowledge in 2026 first appeared on BitcoinWorld.