The Rungs in the Middle: How AI Is Eroding the Pathways, Not Just the Jobs
The AI labor market conversation has a frame problem. It has been organized almost entirely around a single question: which jobs will be destroyed? The question is not wrong. But it is incomplete in a way that matters enormously for the 70 million Americans who do not hold a four-year degree and whose economic lives depend not on surviving in one job but on moving between them.
A new Brookings Institution and Opportunity@Work report, published April 2, 2026, shifts the frame. The question is not just which jobs AI disrupts. The question is which pathways AI erodes — and the answer is: nearly half of them.
The Ladder Is Not the Issue
The standard image of AI displacement is of a worker in a job that gets automated away. That image is not wrong. What it misses is how jobs function for workers who don't have four-year degrees: not as individual positions to be held but as rungs in a sequence that, over years, allows someone to accumulate skills, demonstrate reliability, and advance into higher-wage work.
Opportunity@Work developed the conceptual vocabulary for this. They call workers skilled through alternative routes "STARs" — those who built knowledge through work experience, military service, apprenticeships, or community college rather than a bachelor's degree. There are 70 million of them in the US workforce.
The Brookings report maps how STARs move through three categories of occupations. Origin occupations are accessible entry points: cashiers, couriers, receptionists. Gateway occupations are the critical middle tier: customer service representatives, administrative assistants, payroll clerks. Destination occupations are the higher-wage endpoint: human resources managers, sales representatives, financial analysts.
The key analytical insight is that economic mobility is not produced by any single job. It is produced by the transition between jobs — by the fact that a worker who masters a Gateway role develops transferable skills and a credible employment history that makes the next step possible. Research consistently shows that workers are more likely to advance into higher-paying occupations that share underlying skill similarities with their current role. The pathway is the mechanism.
STARs represent 62.3% of workers in Gateway occupations. This is not incidental. Gateway jobs are designed for people who learn by doing, not by credentialing.
What the Data Shows
The Brookings report applies Anthropic's "observed exposure" measure — an estimate of the share of tasks within an occupation that AI can currently perform or assist with, based on actual deployment data from the Claude large language model — to the Origin-Gateway-Destination framework. The results:
15.6 million STARs work in occupations in the top quartile of AI exposure. That is one-fifth of the entire STAR workforce. Nearly 11 million STARs are in Gateway occupations that are highly AI-exposed. Six Gateway occupations alone account for almost 8 million of those workers, concentrated in clerical, administrative, and customer service roles. Only 51% of pathways between Gateway and Destination occupations are not highly AI-exposed — meaning nearly half of the established routes between middle-tier jobs and higher-wage work are in the zone of significant disruption.
The last number is the one that deserves attention. The debate about AI and jobs has mostly been about whether specific positions survive. The Brookings data suggests the more consequential question is whether the transition pathways survive. A Gateway job that remains nominally intact but has shed the tasks that made it a learning environment — the problem-solving, the human interaction, the judgment calls — is no longer a functional rung on the ladder. It becomes a holding position.
The Scarring Compound
This connects directly to a pattern Goldman Sachs identified in April 2026: workers displaced by technological disruption face long-term economic costs that persist well beyond the initial displacement. Goldman's analysis of 40 years of labor market data finds that technology-displaced workers experience an average 3% cut in real earnings compared to workers in stable occupations, with 10 percentage points less real earnings growth over a decade. The job searches are longer. The risk of future unemployment is higher. Homeownership and wealth accumulation are delayed.
Goldman's chief concern for 2026 is what senior economist Joseph Briggs called "frontloading" — job losses arriving faster than the 10-year baseline transition timeline, which would amplify the aggregate economic shock. The Brookings framework suggests a mechanism by which frontloading gets multiplied: if AI erodes Gateway occupations faster than new pathways develop, workers displaced from Origin roles have nowhere to go. The pipeline does not simply slow. It breaks.
Anthropic's own labor market research, published March 2026, offers supporting data from the entry point. The job-finding rate for workers aged 22-25 entering highly AI-exposed occupations has declined approximately 14% since 2022. This slowdown is specific to new entrants; it does not appear in workers over 25. The most straightforward interpretation: companies are pausing junior hiring in roles where AI can perform the entry-level tasks. The entry point to the Gateway is narrowing.
Three data points, three stages of the same problem. Entry is harder (Anthropic). The middle of the pathway is under pressure (Brookings). Displacement, when it happens, leaves lasting marks (Goldman). This is not three separate labor market stories. It is one story told at three points in a single career arc.
What Exposure Does and Doesn't Tell You
The Brookings report is careful on this point, and it is worth preserving its caution rather than overwriting it. Exposure does not equal displacement. AI can augment a role rather than eliminate it — the report acknowledges scenarios in which AI tools accelerate skill development for STARs, narrow the skill distance between Gateway and Destination roles, and expand access to higher-wage work. Economists Daron Acemoglu, David Autor, and Simon Johnson argue that AI's potential as a collaborator may in many cases enable new tasks and increase the value of human expertise.
That argument is real. The question is: how is AI actually being deployed in the Gateway occupations where 11 million STARs currently work? Anthropic's observed-exposure measure captures actual deployment, not theoretical capability. The 14% hiring decline for young workers in high-exposure roles is not a theoretical projection. It is what the labor market data shows is already happening at the entry point.
The optimistic scenario requires employers to use AI in ways that strengthen transition pathways — that make Gateway roles richer learning environments rather than thinner ones. There is nothing in the current policy landscape that creates systematic incentives for employers to make that choice. The decision about whether AI augments or displaces the middle rungs of the American career ladder is being made, right now, at the level of individual firms, without any particular obligation to consider what happens to the workers downstream.
Who Pays
One note on distribution. The Brookings report finds that the highest rates of AI-related pathway exposure are in administrative, clerical, and customer service Gateway occupations. These roles are disproportionately held by women. They are concentrated in the Northeast and Sun Belt. The workers in them are precisely the workers Brookings describes as having "low adaptive capacity" — limited ability to weather displacement and transition to new work.
The Brookings report estimates that 3.5 million STARs — 67% of workers who are both highly AI-exposed and have low adaptive capacity — face the compounded risk: their current role is highly exposed, and if it disappears, they have limited resources for transition. These are workers in their 40s and 50s who do not own significant financial assets, who cannot easily relocate, who are not candidates for retraining programs that assume high baseline digital literacy.
The distribution of these risks is not random. It follows the existing geography of inequality, landing hardest on the workers who were already the most economically vulnerable and the least represented in the conversations about how AI should be deployed.
The Structural Question
The Brookings report ends with a set of urgent questions for the field. That framing is honest about the limit of what a research paper can do. It cannot create the workforce system investments, the employer incentives, or the regulatory frameworks that would determine whether AI strengthens or severs the pathways 70 million American workers depend on. Those decisions are political.
The economic structure that produced the Origin-Gateway-Destination mobility system was itself built by a combination of labor market pressures, employer practices, and policy choices over forty years. It was not inevitable that customer service work would be a Gateway occupation — that it would be the kind of role that taught people skills transferable to human resources or sales. It became that because of how those jobs were designed, compensated, and organized.
AI is not redesigning that system according to a plan. It is being deployed, role by role, by employers making decisions about cost and efficiency. The aggregate effect of those decisions is a research question. The Brookings report has made a serious start at answering it. The answer so far is not reassuring.
Sources
Brookings Institution / Opportunity@Work. How AI May Reshape Career Pathways to Better Jobs. April 2, 2026. https://www.brookings.edu/articles/how-ai-may-reshape-career-pathways-to-better-jobs/
Anthropic. Labor Market Impacts of AI: A New Measure and Early Evidence. March 2026. https://www.anthropic.com/research/labor-market-impacts
Goldman Sachs Research. Pierfrancesco Mei and Jessica Rindels, "AI Displacement and Labor Market Scarring," April 7, 2026. Internal research note; no public DOI. The 3% real earnings cut and 10 percentage point slower earnings growth figures derive from this report, which analyzed 40 years of longitudinal data tracking more than 20,000 workers displaced by technological change. Reported via Business Insider, April 6, 2026: https://www.businessinsider.com/ai-job-cuts-layoffs-finding-new-job-pay-cut-goldman-2026-4. Separate Goldman Sachs analysis: Joseph Briggs, "How Will AI Affect the US Labor Market?" https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-us-labor-market
Acemoglu, Daron, David Autor, and Simon Johnson. NBER Working Paper 34854. 2026. https://www.nber.org/system/files/working_papers/w34854/w34854.pdf
Opportunity@Work. Navigating with the STARs. https://www.opportunityatwork.org/topics/reports/navigating-with-the-stars
Manning, Sam, and Tomas Aguirre. Measuring US Workers' Capacity to Adapt to AI-Driven Job Displacement. Brookings. https://www.brookings.edu/articles/measuring-us-workers-capacity-to-adapt-to-ai-driven-job-displacement/
