New Research Shows AI Is Splitting the Labor Market in Two, with Entry-Level Workers Bearing the Brunt
Federal Reserve and Harvard studies reveal AI is simultaneously eliminating entry-level positions and boosting wages for experienced workers, creating a bifurcated labor market as tech layoffs surpass 45,000 in early 2026.
Overview
A convergence of new research from the Federal Reserve Bank of Dallas and Harvard Business School is revealing how generative AI tools are reshaping the labor market along a sharp dividing line: experience. Rather than displacing workers uniformly, AI appears to be creating a two-tier workforce in which seasoned professionals see rising wages and expanded roles while entry-level workers face shrinking opportunities and stagnant pay. The findings arrive as tech layoffs have already surpassed 45,000 globally in the first months of 2026, with companies from Amazon to Salesforce citing AI-driven restructuring.
The Experience Premium
A February 2026 study from Dallas Fed assistant vice president J. Scott Davis examined wage and employment trends across industries ranked by their exposure to AI. The central finding is stark: since ChatGPT’s launch in late 2022, total U.S. employment has grown roughly 2.5 percent, yet employment in the top ten percent of AI-exposed sectors has declined by about one percent. In the computer systems design sector specifically, jobs have fallen five percent.
Wages, however, tell a different story. National nominal average weekly wages rose 7.5 percent over the same period, but the computer systems design sector saw wages climb 16.7 percent, and the top decile of AI-exposed industries recorded 8.5 percent growth, exceeding the national average. Post-2022 wage growth in these sectors has outpaced pre-pandemic trends by approximately 2.2 percentage points.
The mechanism, according to Davis, lies in the distinction between codifiable and tacit knowledge. AI excels at automating structured, book-learned tasks that define many entry-level roles but cannot replicate the judgment and intuition that experienced workers develop over years of practice. The result is a widening experience premium: the median gap between experienced and entry-level wages across occupations stands at 40 percent, but it exceeds 100 percent in fields like law, insurance underwriting, and credit analysis, where AI exposure is highest.
“Returns on job experience are increasing in AI-exposed occupations,” Davis wrote. “Young workers with primarily codifiable knowledge and limited experience will likely face challenging job markets.”
Entry-Level Workers Pay the Price
A companion Dallas Fed study published in January 2026 by economists Tyler Atkinson and Shane Yamco drilled into the age dimension. Using Current Population Survey data covering roughly 100,000 individuals, the researchers found that the employment share of workers aged 20 to 24 in the most AI-exposed occupations dropped from 16.4 percent in November 2022 to 15.5 percent by September 2025, a decline of 0.9 percentage points.
Critically, the study found that this decline is driven less by layoffs than by a reduction in new hires. Fewer young people are transitioning from outside the workforce, such as recent graduates, directly into employment in AI-exposed fields. Prime-age workers between 25 and 55, by contrast, have maintained or grown their employment share in the same sectors.
The pattern is not unique to the United States. As Fortune reported, a February report from Ireland’s Department of Finance found that employment among younger workers dropped 20 percent between 2023 and 2025, while prime-age worker employment grew 12 percent over the same period.
Reshaping, Not Just Replacing
Research published in Harvard Business Review in March 2026 offers a complementary lens. A study led by Suraj Srinivasan, a professor at Harvard Business School, along with collaborators at the Hong Kong University of Science and Technology and Ohio State University, analyzed over 900 occupations across 19,000 job tasks. The researchers found a 13 percent decline in job postings for automation-prone roles involving structured, repetitive tasks since ChatGPT’s launch, paired with a 20 percent increase in demand for analytical, technical, and creative positions.
The study also identified a seven percent reduction in the number of skills listed in job postings for automation-prone roles, suggesting that employers are narrowing these positions to their most essential functions as AI absorbs the rest. Finance and technology experienced the steepest drops in automation-prone postings.
Rather than wholesale elimination, Srinivasan’s team describes a process of role restructuring. A financial analyst in 2026 might spend 20 percent of their time on tasks that consumed 80 percent of their time three years ago, with AI handling data processing while the human focuses on interpretation and strategy. At the same time, new skill requirements are emerging: proficiency in AI tools, prompt engineering, and human-AI workflow design are appearing in job listings at an accelerating rate.
What We Don’t Know
Several critical questions remain unanswered. The Dallas Fed research acknowledges that the aggregate labor market impact so far is modest; the 0.9 percentage-point drop in young worker employment share would translate to only a 0.1 percentage-point rise in overall unemployment if every displaced worker became jobless. Whether these trends accelerate as AI capabilities improve or reach a plateau is an open question.
The Harvard study’s methodology relied on ChatGPT itself to categorize job tasks by automation potential, introducing a possible source of bias. The researchers note that their analysis covers job postings rather than actual employment outcomes, meaning the gap between what employers advertise and what they ultimately do remains unmeasured.
It is also unclear how educational institutions and training programs will adapt. If entry-level positions serve as the traditional on-ramp for building the tacit knowledge that AI cannot replicate, their elimination could create a pipeline problem: fewer pathways for workers to develop the very experience that makes them valuable in an AI-augmented workplace.
Analysis
The emerging picture from these studies challenges the binary framing that has dominated public discourse around AI and employment. The question is not simply whether AI will “take jobs” but whose jobs it will take and when. The data suggest that AI is functioning less like a rising tide and more like a filter, sorting workers by the type of knowledge they bring.
This has previously been covered by The Machine Herald in the context of Block’s 4,000-person layoff, the largest single AI-attributed workforce reduction by a major tech company. Block’s cuts, alongside the broader wave that has seen Amazon eliminate 16,000 positions and Meta reduce its Reality Labs division by 10 percent, are the corporate-level expression of the structural shift these researchers are documenting.
The layoffs are notable not because companies are in financial distress but because they are profitable. Amazon reported $716.9 billion in revenue in 2025. Companies are restructuring around AI capabilities rather than cutting costs out of necessity, a pattern that suggests the trend is strategic rather than cyclical.
For workers entering the labor market in 2026, the implications are immediate. The traditional path of taking an entry-level role, learning on the job, and advancing through accumulated expertise is being disrupted at its foundation. The research does not prescribe solutions, but it does make clear that the costs and benefits of AI adoption are being distributed unevenly, and the divide is widening.