Is AI actually taking entry-level jobs? Here's what the data says

On the evidence available as of August 2026, AI is not gutting entry-level hiring across the economy, but it is measurably thinning it in the roles it can substitute for. The Stanford Digital Economy Lab's payroll-data study found employment for workers aged 22-25 in AI-exposed occupations running 19% below where it would sit had it kept pace with less-exposed peers — a gap that has widened since August 2025, not a sudden cliff.

That finding comes from the Stanford Digital Economy Lab's payroll-data research, an August 12, 2026 update to a running analysis of millions of U.S. workers' ADP payroll records through June 2026, led by economist Erik Brynjolfsson with Bharat Chandar and Ruyu Chen. NPR's August 18, 2026 reporting on that research adds a second layer: what recent graduates themselves believe is happening, and where economists disagree with them. A third source, Gallup's July 21, 2026 workplace report, covers a related but separate question — what AI use does to the engagement of people who already have a job.

What does the payroll data actually show?

Young workers in occupations most exposed to generative AI are seeing hiring slow, not mass layoffs. The Stanford Digital Economy Lab found no economy-wide job displacement, but a persistent 19-point gap for 22-to-25-year-olds in exposed roles, driven mainly by reduced hiring rather than firing.

The study tracks ADP payroll records rather than surveys, which is why it can separate hiring slowdowns from layoffs. Roles where AI substitutes for tasks show flat or falling entry-level headcount; roles where AI complements the work show flat or rising headcount. Wages, notably, have barely moved — the adjustment shows up almost entirely in who gets hired, not in what anyone gets paid.

So is AI actually the cause?

Economists interviewed by NPR are split, and neither side dismisses the data — they dispute what's driving it. Brynjolfsson calls AI "part of the story," while Harvard's David Deming points to hiring softness that predates ChatGPT's public release, arguing remote work is the bigger factor.

Deming's case, per NPR, rests on timing: junior hiring at many firms declined roughly six months before generative AI tools were widely available, and a New York Fed analysis he cites points to remote work making it costlier to train juniors as a competing explanation. Brynjolfsson's counter is structural: entry-level workers get cut first in downturns, and their tasks overlap more with what AI models were trained on than a senior worker's tacit, undocumented knowledge does.

Perception is running ahead of the data either way. A ZipRecruiter survey cited in NPR's reporting found 47% of 2026 graduates already believe AI has affected hiring in their field — a belief the underlying payroll numbers only partly support, and one none of the three economists NPR spoke with treats as settled fact.

Don't companies that use AI heavily hire more juniors, not fewer?

Some do. University of Chicago economist Anders Humlum, drawing on a Ramp and Revelio Labs study of more than 21,000 firms cited in NPR's report, found companies with the heaviest AI investment grew entry-level headcount by 12% over two years after adopting the tools — the opposite of the Stanford-level pattern.

The two findings aren't a contradiction so much as two different aggregation levels. Stanford's data captures occupation-level exposure across the whole economy; Humlum's captures company-level behavior among firms that chose to invest heavily in AI, which may be growing overall and hiring more of everyone, juniors included. Both can be true at once — which is itself the honest, unsatisfying answer.

Source	What it measured	Headline finding	As of
Stanford Digital Economy Lab	ADP payroll data, economy-wide	22-25-year-olds in AI-exposed roles: employment 19% below trend	Aug 12, 2026
Ramp / Revelio Labs (via NPR)	21,000+ firms, company-level	Heaviest AI adopters grew entry-level headcount 12% over two years	Reported Aug 18, 2026
ZipRecruiter (via NPR)	Survey of 2026 graduates	47% believe AI has already affected hiring in their field	Reported Aug 18, 2026

Does using AI at work actually make people more engaged, or just busier?

For people who already have a job, Gallup's July 2026 data shows AI use correlates with higher engagement, but only when management backs it. Weekly AI users were 8 points more engaged than infrequent users, and engagement hit 53% when frequent use, a clear rollout plan, and manager support were all present together.

Without that support, the gains shrink fast: engagement ran 48% with manager backing versus 30% without it, on Gallup's quarterly survey of over 43,000 U.S. employees fielded in February and May 2026. National engagement overall stayed flat at 31%, down five points from 2020's peak — AI adoption is accelerating, in other words, but it isn't on its own reversing the broader engagement slide Gallup has tracked for years.

What does this mean at your desk?

If you're early in your career in a role AI can substitute for — drafting, summarizing, first-pass coding — the honest read of the data is a slower hiring market for people like you, not a jobless one, and worth planning around rather than panicking over. If you manage juniors or hiring, the Humlum data is the more useful signal: firms investing seriously in AI aren't uniformly cutting entry-level roles, and a documented rollout plan with active manager support is the difference Gallup's numbers put a real number on.

What none of these three sources establish is a single, economy-wide verdict on AI and jobs. Stanford's own team frames the pattern as an early signal, not a forecast, and the two Harvard and Stanford economists NPR interviewed still disagree on how much of it is AI at all.

FAQ

Is AI really replacing junior employees right now? Not broadly. Stanford's payroll-data research found no economy-wide job displacement as of its August 2026 update — the effect is a hiring slowdown concentrated in AI-exposed roles for workers aged 22-25, not mass layoffs.

Which jobs are most affected? Per the Stanford research, roles where AI tools substitute directly for the work — drafting, first-pass analysis, routine coding — show flat or falling entry-level hiring, while roles where AI complements human work show flat or rising headcount.

Do companies that adopt AI heavily hire fewer new grads? Not necessarily. The Ramp/Revelio Labs study of 21,000-plus firms, cited in NPR's reporting, found the heaviest AI adopters grew entry-level headcount by 12% over two years — the opposite pattern from the economy-wide occupation data.

Does using AI at work make employees happier? Gallup's July 2026 data ties AI use to higher engagement, but the gain depends heavily on manager support — 48% engagement with it versus 30% without, among frequent AI users surveyed.

For a related workflow perspective, read What ChatGPT, Claude, and Gemini actually remember about you, per their own documentation.