--- title: "Workers Aged 22 to 25 in AI-Exposed Jobs Saw Employment…" canonical: "https://www.metaintro.com/blog/dallas-fed-young-workers-ai-exposed-jobs-employment-decline" language: "en" author: "drashtigarach" published: "2026-09-02T10:39:37.000Z" modified: "2026-09-02T11:39:57.194Z" --- [Back to Blog](/blog) [AI](/blog/tag/ai)[Jobs](/blog/tag/jobs)[Labor Market](/blog/tag/labor-market)[Education](/blog/tag/education) # Workers Aged 22 to 25 in AI-Exposed Jobs Saw Employment Fall 13 Percent Since 2022 Dallas Fed research finds employment down 13 percent for workers aged 22 to 25 in AI-exposed jobs since 2022, and the cause is blocked entry, not layoffs. [![Drashti Garach](https://cdn.metaintro.com/rs:fill:40:40/q:72/plain/images/5719d740-e510-42bc-8017-e040d145f35f_1766029465094.png)Drashti Garach @DrashtiGarach](/blog/author/drashtigarach) [September 2, 2026](/blog/archive/2026/09)12 min read ![A college computer training lab before opening, screens off](https://cdn.metaintro.com/rs:fill:1200:675/q:78/plain/images/10117e-computerlab-hero.png) [https://x.com/intent/tweet?text=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline](https://x.com/intent/tweet?text=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline&title=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline&title=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022)[mailto:?subject=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline](mailto:?subject=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline) Research from the [Federal Reserve Bank of Dallas](https://www.dallasfed.org/) finds that workers aged 22 to 25 in the most AI-exposed occupations have seen employment fall 13 percent since 2022 while employment for older or less exposed workers stayed steady or rose, a finding the [Financial Times](https://www.ft.com/content/9334212a-3cc4-426c-9cbe-57cb48033603) picked up in its reporting on graduates in the heart of the data centre boom. At [Metaintro](https://www.metaintro.com), we care less about whether AI is coming for jobs in general and more about which specific rung of which specific ladder is moving, and this research answers that unusually precisely. The damage is concentrated at the entry point, it is not showing up as layoffs, and it barely registers in the national unemployment rate. ## What did the Dallas Fed actually find? The [January 2026 analysis](https://www.dallasfed.org/research/economics/2026/0106) reports a correlation across occupations between employment declines and AI exposure, but only for younger workers, which it says suggests only a slight impact on the aggregate unemployment rate so far. The headline 13 percent decline for workers aged 22 to 25 in the most exposed occupations comes from a [Stanford University](https://digitaleconomy.stanford.edu/) study built on payroll records covering 25 million workers held by [ADP](https://www.adp.com/), scoring occupations for AI exposure and sorting them into quintiles. The Dallas Fed then tested that finding against different data. Rather than proprietary payroll records it used public microdata from the [Census Bureau's Current Population Survey](https://www.census.gov/programs-surveys/cps.html), a rotating panel of around 100,000 individuals, which is a smaller sample but reaches further back in time and tracks people moving into and out of occupations. Measured that way, the share of employment held by workers aged 20 to 24 in the most AI-exposed occupations slipped from 16.4 percent in November 2022, when ChatGPT was released, to 15.5 percent in September 2025. The two datasets disagreeing slightly is a feature rather than a problem. Payroll records capture people on a company's books in near real time but only cover firms using that provider, while a household survey captures anyone in the labor force but with more statistical noise once you slice it by age and occupation together. The Dallas Fed flags that limitation openly, noting that partitioning an already small sample by age and AI exposure leaves relatively few observations in any one group. Two independent methods pointing the same direction is stronger evidence than either alone, and it is also why the precise size of the effect should be held loosely while the direction should not. ## Is a 0.9 percentage point drop actually a big deal? Not for the national statistics, and the Dallas Fed says so plainly. Workers aged 20 to 24 make up only about 9 percent of the labor force, so a 0.9 percentage point decline in that group's share of AI-exposed employment likely explains little of the rise in the aggregate unemployment rate. Anyone using this research to argue that AI is driving national joblessness is overreading it, and the same caution runs through [the finding that the youth unemployment jump came from frozen hiring rather than AI](https://www.metaintro.com/blog/gen-z-job-market-not-ai-fault). For an individual it is a different calculation entirely. National averages describe a labor market, not a career, and if you are 23 and applying into one of those occupations you are competing inside the group where the effect is concentrated rather than inside the aggregate where it disappears. Both things are true at once, which is why [one recent study arguing AI is not why young workers cannot find jobs](https://www.metaintro.com/blog/ai-isnt-why-young-workers-cant-find-jobs-study) and this one are less contradictory than they look. The dispute is about magnitude and mechanism, not about whether the entry rung got harder. ## Why does it matter that this is not layoffs? Because it changes what you do about it. The Dallas Fed found no increase in young workers in high AI exposure occupations moving from employment into unemployment, and no distinctive pattern in job-finding rates among those already unemployed. What it found instead was a decline in people transitioning directly from outside the workforce into employment. In plain terms, the people already in these jobs are keeping them. The people trying to get in are not getting in. That is a hiring-gate problem rather than a job-security problem, and the two demand opposite responses. If this were a layoff wave, the advice would be about severance, notice periods and moving fast. Because it is a blocked entry point, the advice is about how you get considered at all, which is the same structural issue behind [the first rung vanishing and how new graduates break in without entry roles](https://www.metaintro.com/blog/how-2026-grads-break-in-ai-entry-level-jobs-vanishing). Anyone already employed in an exposed occupation is in a considerably stronger position than the headlines suggest. It also explains a frustration a lot of early-career candidates describe, which is doing everything right and still hearing nothing back. When the constraint is that a rung has been removed rather than that a bar has been raised, additional qualifications do not unlock the door, because the door is not being guarded, it is missing. That distinction is worth internalising before you spend money on another certificate. The spending that pays in this situation is whatever puts you in contact with a person who can create or fill a role, not whatever adds another line to a resume that nobody is reading. ## Which jobs count as AI-exposed? The Dallas Fed's [Texas-focused work](https://www.dallasfed.org/cd/communities/2025/2505) is the clearest guide, and it drew on the Department of Labor's [occupational skills database](https://www.onetonline.org/) to compare jobs by the skills and work activities they involve. Jobs built around writing, gathering information and computer programming score as more exposed to AI change. Jobs built around operating machinery, or work in the trades, score as less exposed. That framing is more useful than a list of job titles because it works on tasks. Two people with the same title can have very different exposure depending on what they actually do all day, and the sensible exercise is to break your own week into its component activities and score them honestly rather than to look up your occupation and accept the verdict. It also explains why demand keeps rising in physical work even as office entry roles tighten, a split visible in [three skilled trade openings for every qualified worker](https://www.metaintro.com/blog/skilled-trade-openings-per-qualified-worker-2026). The Dallas Fed is careful to note that AI-prompted change can be positive or negative for a worker depending on how it lands on the job. The Texas research also makes a point that gets lost in coverage aimed at graduates. It looked specifically at so-called good jobs, meaning roles that do not require a bachelor's degree but still pay a self-sustaining wage, because those are the jobs training programmes target and the ones that carry people into economic security without a degree. Finding that machinery and trades work scores as less exposed is therefore not a consolation prize. It is a finding that one of the main routes to a stable income for people without degrees looks more durable under AI than the white-collar route that graduates were told to take. ## Is AI replacing workers or making them more valuable? Both, and which one you get depends on the kind of knowledge your job runs on. The Dallas Fed's [February 2026 analysis](https://www.dallasfed.org/research/economics/2026/0224) draws the distinction between codified knowledge, the sort available in a textbook, and tacit knowledge, the sort you only acquire by doing the work. If AI can reproduce codified knowledge but not tacit knowledge, it substitutes for the roles built on the former and complements the roles built on the latter. The employment data fits that split. Total US employment rose about 2.5 percent after ChatGPT's release in late 2022, but employment in computer systems design and related services fell 5 percent, and across the 10 percent of sectors most exposed to AI it fell about 1 percent. Meanwhile wages are rising in AI-exposed occupations that place a high value on experience and tacit knowledge. So the same technology is pushing down the number of entry roles while pushing up the price of the people who already have judgment, which is a fairly brutal description of why [the AI skills gap shows up directly in salaries](https://www.metaintro.com/blog/ai-skills-gap-and-your-salary). The uncomfortable implication is generational rather than technical. A market that pays more for tacit knowledge while creating fewer opportunities to acquire it is one where the people who already have experience capture the gains and the people who need to build it face a narrower path. That is not a permanent state, because employers eventually need to replace the experienced people, but it can persist for several years, and several years is a long time at the start of a working life. ## What does this mean for your career? If you are early in your career and targeting an exposed occupation, the binding constraint is being considered, not being qualified. That argues for evidence over credentials. A small portfolio of finished work, a referral into a specific team, or a contract or internship that converts will move you further than another application into a queue, because the mechanism blocking you is the absence of an open entry rung rather than a judgement about your ability. Widening your target list to adjacent, less exposed occupations that use the same underlying skills is the other lever, and it is usually faster than waiting for the rung to reappear. If you are already established, the research is close to reassuring. Employment for more experienced workers has been steady or increasing, wages are rising where experience is valued, and there is no evidence of an AI-driven layoff wave in these numbers. The risk for you is slower and less dramatic, which is that the tacit knowledge premium only holds while your judgement stays ahead of what the tools can do. "In combination with layoffs over the recent years plus the rise of required AI skills, experience is no longer enough," Lacey Kaelani, chief executive of [Metaintro](https://www.metaintro.com), told [Fortune](https://fortune.com/2025/12/16/job-market-so-bad-people-age-40-going-back-to-school/). The Dallas Fed data suggests experience is still worth a great deal, and that keeping it current is the whole job. ## What should employers take from this? That they are quietly dismantling their own pipeline. If entry-level hiring in exposed occupations is being cut because AI can do some of what a junior used to do, the organisation is trading a short-term cost saving against the supply of mid-level people it will need in five years, since tacit knowledge is acquired by doing the work and there is no other route to it. That trade rarely appears in the business case, and it is the substance of [why cutting entry-level jobs could backfire on employers](https://www.metaintro.com/blog/why-cutting-entry-level-jobs-could-backfire-on-employers). There is a practical version of the fix that does not require reversing the automation. Redesign the junior role around the tasks AI cannot do, which in this framing means putting new people closer to judgement, client contact and ambiguity earlier than tradition allows, rather than having them serve an apprenticeship on the codifiable work that has now been automated. Analysis from the [New York Fed](https://libertystreeteconomics.newyorkfed.org/2026/05/do-job-postings-show-early-labor-market-effects-of-ai/) on what job postings reveal about early AI effects, and from [Stanford's economic policy researchers](https://siepr.stanford.edu/publications/policy-brief/what-really-happening-jobs-separating-ai-hype-reality) on separating the hype from the measurable, are both worth reading before making that call on the basis of a headline. ## Related Articles - [Entry Level Hiring Just Halved, How New Graduates Can Still Land a Job](https://www.metaintro.com/blog/uk-graduate-vacancies-halved-2026-entry-level-hiring) - [What Business Leaders Get Wrong About AI and Entry-Level Jobs](https://www.metaintro.com/blog/ai-entry-level-jobs-leaders-2026) - [PwC Cuts Advisory Hiring as Entry-Level Consulting Shrinks](https://www.metaintro.com/blog/pwc-advisory-offices-72-to-13-entry-level-hiring) - [How to Upskill for AI Without Going Back to School](https://www.metaintro.com/blog/how-to-upskill-for-ai-without-going-back-to-school) - [Best Jobs for a Career Change and Why Transferable Skills Matter More](https://www.metaintro.com/blog/best-jobs-for-a-career-change) - [Why Gen Z Workers Are Pushing Back on AI at Work](https://www.metaintro.com/blog/gen-z-pushback-ai-at-work-2026) - [How to Make a Resume for Your First Job With No Work Experience](https://www.metaintro.com/blog/how-to-make-a-resume-for-your-first-job) - [Why a Career Passport Could Replace the Resume for Graduates](https://www.metaintro.com/blog/career-passport-replace-resume-2026-graduates) - [The Reskilling Fix for the AI Skills Apocalypse](https://www.metaintro.com/blog/reskilling-fix-ai-skills-apocalypse-flight-simulator) - [Data Center Boom Has Employers Paying Off Student Loans to Land Electricians](https://www.metaintro.com/blog/data-center-boom-skilled-trades-hiring-perks-2026) ## People Also Asked ### Q: Is AI actually taking entry-level jobs? A: The evidence points to blocked entry rather than replacement. Dallas Fed research found employment down 13 percent since 2022 for workers aged 22 to 25 in the most AI-exposed occupations, driven by fewer people moving from outside the workforce into those jobs rather than by layoffs. Employment for older and less exposed workers stayed steady or rose over the same period. ### Q: Which occupations are most exposed to AI? A: Work built around writing, gathering information and computer programming scores as more exposed, while work involving operating machinery or the skilled trades scores as less exposed. Exposure is measured by tasks rather than job titles, so two people sharing a title can face very different exposure depending on what their week actually contains. ### Q: Does this mean the unemployment rate is rising because of AI? A: No, and the Dallas Fed says so directly. Workers aged 20 to 24 are only about 9 percent of the labor force, so the roughly 0.9 percentage point fall in their share of AI-exposed employment explains little of the change in the national unemployment rate. The effect is real but narrow, and it is concentrated at the entry point. If the entry rung in your target field is thinner than it was, the answer is usually a different door rather than a different career. [Metaintro](https://www.metaintro.com) indexes live roles across exposed and less exposed occupations alike, so you can see which employers are still opening junior positions instead of inferring it from research. [Create your free profile](https://www.metaintro.com/signup) and find the teams still hiring at your level. ### Share this article [https://x.com/intent/tweet?text=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline](https://x.com/intent/tweet?text=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline)[http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline](http://www.facebook.com/sharer.php?u=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline)[https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline&title=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline&title=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022)[mailto:?subject=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline](mailto:?subject=Workers%20Aged%2022%20to%2025%20in%20AI-Exposed%20Jobs%20Saw%20Employment%20Fall%2013%20Percent%20Since%202022&body=https%3A%2F%2Fwww.metaintro.com%2Fblog%2Fdallas-fed-young-workers-ai-exposed-jobs-employment-decline) For job seekers ## The job you want exists. We’ll find it. No more scrolling job boards. Get matched roles by text, with the pay and the reason they fit. [Get Started Free](/signup)[Search matching jobs](/jobs/search) - MatchCompare live roles against your current evidence. - PositionTurn proof projects into role-specific applications. - ImproveUse market feedback to keep the skill plan current. [Return to navigation](#main-navigation)