AI Won't Kill Your Job. It Already Killed the Path That Was Supposed to Make You Good At It.

The economists who told us for three years that AI would create as many jobs as it destroys have quietly started updating their models. The data coming out of the Dallas Federal Reserve, Yale School of Management, and Stanford University in 2026 tells a story they didn't anticipate: AI isn't destroying jobs at the economy level. It's destroying something harder to measure and more consequential , the entry points into careers.

What Actually Happened

Three significant research outputs landed in early 2026 that collectively reframe the AI employment debate. The Federal Reserve Bank of Dallas found that employment in the 10 percent of sectors most exposed to AI has declined 1 percent since ChatGPT's November 2022 release, while total U.S. employment grew approximately 2.5 percent over the same period , a 3.5 percentage point divergence. Computer systems design , the sector encompassing most software development roles , has seen employment fall 5 percent since late 2022. Crucially, the decline falls almost entirely on workers under age 25; employment totals for older workers in the same sectors have not declined.

Yale's Jeffrey Sonnenfeld and colleagues found that 58 percent of Gen Z graduates from 2024 and 2025 were still looking for their first job in early 2026, compared to just 25 percent of millennial and Gen X graduates at equivalent career stages. Meanwhile, Goldman Sachs research published in April 2026 estimates AI is eliminating approximately 16,000 U.S. jobs per month , but the distribution is strikingly uneven. Employment among developers aged 22 to 25 has fallen nearly 20 percent from its late-2022 peak, while experienced workers in the same sectors have seen wages rise 16.7 percent , more than double the 7.5 percent national average wage growth over the same period.

Why This Matters More Than People Think

The headline statistics matter less than the mechanism the Dallas Fed identified. Federal Reserve economist Tyler Atkinson is explicit: the decline in young worker employment in AI-exposed sectors is not being driven by layoffs. Experienced workers are not being pushed out. Instead, the decline is driven by a low job-finding rate , people leaving education cannot find positions to enter. The gate is closed, not the exit. That distinction changes everything about the downstream implications.

Layoffs are visible, politically legible, and generate regulatory pressure. Closed entry points are invisible to most policy frameworks. You don't file for unemployment when you can't get hired in the first place. You are simply counted among the "not in labor force" or "unemployed" statistics without generating the political urgency that a mass layoff event would produce. ServiceNow CEO Bill McDermott warned in March 2026 that youth unemployment could hit 30 percent if the current trajectory continues , a level that would exceed the United States' own Great Depression peak of 24.9 percent. Forty-one percent of university leaders surveyed before a Yale Higher Education Summit reported being highly concerned about the vulnerability of entry-level white-collar roles.

The Competitive Landscape

The companies benefiting from eliminating entry-level roles are not unaware of the dynamic. In fact, they are actively marketing it as an efficiency gain. Coinbase laid off 700 employees , 14 percent of its workforce , in May 2026, explicitly framing the restructuring around moving to "AI-native pods" where smaller teams of senior engineers supported by AI agents accomplish what previously required much larger headcounts. Snap laid off 1,000 employees in April 2026, with internal communications noting that 65 percent of coding tasks are now handled by AI automation. In both cases, the cuts concentrated on junior and mid-level roles.

The perverse irony is that the companies leading AI adoption are simultaneously reducing the pipeline of human workers who will understand how to manage, audit, and correct AI systems in the future. Junior employees don't just do entry-level work , they develop the judgment that makes them valuable senior professionals. The Dallas Fed researchers noted that the employment decline falls disproportionately on workers under 25 precisely because those workers perform "codified knowledge" tasks , work that can be taught from textbooks and therefore automated. The work that cannot yet be automated is "tacit knowledge" , the intuition that develops from years of making mistakes and iterating with senior colleagues. That knowledge requires an apprenticeship pipeline to exist. The pipeline is closing.

Hidden Insight: The Knowledge Transfer Crisis Nobody Is Modeling

Here is the second-order consequence almost entirely absent from current policy discussions: if the entry-level jobs that develop junior workers into senior workers disappear, where do the senior workers of 2035 come from? Every experienced engineer, analyst, doctor, lawyer, and financial professional working today developed their expertise through years of lower-stakes work where they could make mistakes cheaply. The 45-year-old software architect commanding $400,000 in total compensation became that person by being a 22-year-old who wrote buggy code, received code review feedback, and iterated for a decade. Remove the 22-year-old position, and you do not simply eliminate one job , you eliminate the formation process for the 45-year-old.

This is the knowledge transfer crisis. No one is modeling it seriously in current AI impact frameworks, because most economic models treat "skills" as something individuals accumulate through exogenous education , you learn programming and then you apply it. But senior expertise isn't accumulated from textbooks. It's accumulated from years of supervised practice in environments that tolerate error. Junior jobs are, in economic terms, an investment in future senior human capital. Corporations have spent three years discovering that AI can replace the output of junior workers at a fraction of the cost. Almost none of them have modeled what happens to their senior talent pipeline in eight to ten years when there are no juniors to promote.

The historical precedent most relevant here is manufacturing. Between 1980 and 2000, U.S. manufacturing eliminated roughly 3 million junior and apprentice-level positions through automation and offshoring. By 2010, manufacturers were reporting an acute shortage of skilled machinists, welders, and precision operators , the very workers whose career path had been the apprentice positions that no longer existed. The "skills gap" crisis in manufacturing that dominated policy conversations throughout the 2010s was the downstream consequence of a pipeline that closed in the 1980s and 1990s. AI's impact on white-collar work may be producing an identical dynamic , but compressed into a 5 to 7 year cycle rather than 20 years, and affecting knowledge work rather than manual skills.

What to Watch Next

Watch for graduate employment data from the National Association of Colleges and Employers (NACE) in their Fall 2026 survey, which will give the clearest statistical picture of the class of 2026's hiring outcomes. If the 58 percent joblessness figure among 2024 2025 graduates appears again in the 2026 cohort, the trend is structural rather than transitional. Also watch for any announcement from major tech employers , Google, Meta, Microsoft, Amazon , around "early career" or "university hiring" programs. Reductions in or elimination of these programs would signal that the largest employers have definitively decided AI-assisted senior workers are more cost-effective than building a junior talent pipeline.

The most important leading indicator will be law school and MBA enrollment data for the fall 2026 cohort. When entry-level professional services employment dries up, one traditional response is to extend education , stay in school longer, acquire more credentials, delay entry into a hostile job market. If law school applications rise 15 percent or more from the 2025 baseline, it will confirm that a generation of workers is extending their education runway rather than simply failing to find work. The 41 percent of university leaders already reporting high concern about entry-level white-collar roles know this crisis is forming. The question is whether corporations and policymakers will recognize it before the knowledge pipeline they have quietly eliminated becomes impossible to rebuild.

AI isn't taking your job , it already took the job that would have taught you how to do yours.

---