How AI Will Separate the Best from the Rest

At a high-profile summit in Paris on February 10th and 11th, tech leaders competed to make the boldest proclamations about artificial intelligence. "AI will be the most profound shift of our lifetimes," declared Sundar Pichai, CEO of Alphabet. Dario Amodei, chief executive of Anthropic, went even further, calling it "the largest change to the global labor market in human history." OpenAI's Sam Altman, in a blog post, envisioned a future where, "In a decade, perhaps everyone on Earth will be capable of accomplishing more than today’s most impactful person."

Altman’s prediction aligns with an early school of thought that AI would democratize opportunity, benefiting lower-skilled workers the most. When large language models first gained popularity in the early 2020s, economists and business leaders saw them as tools that could level the playing field. Nvidia’s CEO, Jensen Huang, envisioned a world where employees would all become "CEOs of AI agents," empowered by technology that could handle complex tasks such as protein-folding and poetry-writing.

A Widening Divide

Yet, recent evidence challenges this optimistic outlook. AI appears more likely to amplify existing skill gaps rather than bridge them. High performers—those already excelling in complex tasks like research, management, and strategic decision-making—are best positioned to leverage AI’s capabilities. Evaluating AI-generated outputs requires expertise and discernment. Rather than reducing disparities, AI is poised to widen them, much like past technological revolutions.

Early studies did lend credence to AI’s potential as an equalizer. Research from 2023 by Erik Brynjolfsson of Stanford University and Danielle Li and Lindsey Raymond of MIT found that generative AI tools boosted productivity by 34% for novice customer-support workers, enabling them to resolve queries faster and more effectively. By contrast, experienced workers saw little benefit, as the AI merely reinforced their existing methods. This suggested that AI could help lower-skilled workers adopt best practices, narrowing performance gaps.

A similar trend was observed in other fields. MIT researchers Shakked Noy and Whitney Zhang found that weaker writers saw the greatest improvements in their work quality when using OpenAI’s ChatGPT to draft reports and press releases. In some cases, simply using the AI’s unedited output improved results. Likewise, a study led by Jonathan Choi at the University of Southern California showed that AI-enhanced legal work benefited the least-skilled law students the most.

The Commoditization of Skills

However, a more profound effect is overshadowing these gains. Jobs consist of various tasks, which AI either enhances or automates. When technology augments work, as in the case of air traffic controllers—where AI processes flight data but leaves final decisions to humans—wages remain high. But when technology simplifies roles, as self-checkout systems have done for cashiers, skill requirements decline, leading to wage stagnation.

For customer service agents and other low-skilled workers, AI’s long-term impact may be more akin to that of self-checkout systems than air traffic control. Amit Zavery of ServiceNow, a business software company, estimates that AI can now handle over 85% of customer-service cases for some clients without human intervention—a figure that is likely to grow. Initially, AI may boost productivity, but over time, it will automate tasks and commoditize skills, reducing the demand for human workers.

Unlike previous waves of automation, which primarily replaced routine jobs such as bookkeeping and assembly-line work, AI extends into non-routine and creative tasks. It can learn implicitly, recognize patterns, and make predictions without explicit instruction. Eventually, it may be able to generate compelling scripts or design innovative products. For now, it is junior professionals in high-wage industries who appear most vulnerable. At A&O Shearman, a law firm, AI tools already handle much of the routine work once performed by associates and paralegals. Their AI system can analyze contracts, compare them with past deals, and suggest revisions in under 30 seconds. "Top performers are best at using AI for strategic decision-making," notes David Wakeling, the firm’s head of AI.

The Rising Premium on Expertise

Recent economic research supports this view. While early studies suggested that lower performers could benefit simply by copying AI-generated outputs, newer findings reveal that AI is most beneficial in complex domains requiring judgment and experience. In scientific research, business strategy, and financial decision-making, high performers gain far more than their lower-performing peers. In some cases, struggling workers see no improvement—or even fall further behind.

MIT’s Aidan Toner-Rodgers found that an AI tool designed to aid materials discovery nearly doubled the productivity of top researchers while having no measurable impact on those in the bottom third. Elite scientists, equipped with deep subject expertise, could identify promising AI-generated suggestions while filtering out irrelevant ones. By contrast, less experienced researchers struggled to distinguish useful outputs from misleading ones.

Similar patterns have emerged elsewhere. Nicholas Otis of the University of California, Berkeley, found that AI-assisted Kenyan entrepreneurs saw divergent outcomes: stronger entrepreneurs increased profits by over 15%, while weaker ones saw their earnings decline. The key difference was how they applied AI recommendations. Low performers tended to follow generic advice, such as increasing advertising, while high performers used AI to devise tailored strategies—such as securing alternative power sources during blackouts.

In financial markets, Alex Kim of the University of Chicago conducted an experiment where participants used AI to analyze earnings-call transcripts before allocating $1,000 in a simulated investment portfolio. Sophisticated investors increased their returns by nearly 10% with AI assistance, while less experienced investors saw only a 2% gain. The best performers used AI insights to evaluate factors like R&D spending and share repurchases, whereas novices struggled to extract meaningful advantages.

The Future of AI-Augmented Work

As AI reshapes industries, new roles are emerging. Rajeev Rajan of Atlassian notes that AI tools free up a few hours each week for engineers, allowing them to focus on more creative tasks. Junior lawyers now spend less time on routine contract reviews and more time interacting with clients. "Really smart people who may have been bored analyzing earnings releases will benefit the most," says an executive at a major investment firm. "The skill that will be rewarded most in the short term is the imagination to find innovative ways to use AI."

Labor markets have always been shaped by technological upheaval. MIT economist David Autor estimates that around 60% of jobs in America in 2018 did not exist in 1940. Occupations like "airplane designer" emerged in the 1950s, while "conference planner" became a recognized profession in the 1990s. But history suggests that technological change disproportionately benefits skilled workers. During the Industrial Revolution, engineers who mastered new machinery saw their wages soar while unskilled laborers suffered. The computer age rewarded software engineers while displacing typists. AI appears set to follow the same trajectory, favoring those with the judgment and adaptability to navigate increasingly complex environments.

The CEO Effect

Today’s AI tools represent only the beginning. As the technology advances, semi-autonomous AI agents capable of acting independently—like those envisioned by Nvidia’s Huang—may transform workplaces even further. In one sense, AI could make every worker a "CEO of AI agents," as Huang predicts. But the hierarchy will remain. The most talented will still make the best CEOs.