If AI Does the Entry-Level Work, Who Trains the Next Generation of Experts?

The first year of most professional careers has never been glamorous. Junior lawyers review contracts line by line. Junior consultants build the slide decks and financial models a partner will present as their own analysis. Junior software engineers fix small bugs in code they did not write, absorbing the codebase one ticket at a time. None of this work is what anyone signed up to do, and almost none of it is what those professionals will still be doing five years later. It has, historically, been the price of admission, the unglamorous repetition through which expertise is actually built.

AI has become genuinely capable at a meaningful share of exactly that work, and entry-level hiring data over the past three years shows the effect clearly.

The Numbers Behind the Concern

Entry-level job postings in the United States fell by roughly 35 percent between early 2023 and mid-2025, according to labor market research firm Revelio Labs, with declines considerably steeper in some technical roles. A Stanford Digital Economy Lab working paper released in November 2025 found that workers aged 22 to 25 in occupations most exposed to generative AI experienced a 16 percent relative decline in employment following the technology’s spread, a disparity not observed to the same degree among older, more experienced workers in the same fields. A separate Harvard-affiliated study reported in February 2026 found that companies adopting generative AI saw junior employment fall 9 to 10 percent while senior employment remained essentially stable, with entry-level hiring at some AI-adopting companies dropping as much as 80 percent in a given quarter.

The pattern shows up by name as well as by statistic. PwC’s UK arm has explicitly cited AI as a factor in cutting roughly 200 entry-level roles as generative AI reshaped how the firm’s early-career work gets done. Nearly 43 percent of college graduates aged 22 to 27 were underemployed as of December 2025, the highest rate since the pandemic, according to reporting that tracked graduate labor market outcomes through the period.

Why This Work Mattered Beyond Its Output

The uncomfortable part of this pattern is not simply that fewer junior professionals are being hired. It is that the specific work being automated away was never really about the work product itself. A partner did not need a first-year associate to review a contract because no one more experienced was available; the review was, in large part, how that associate learned to spot the clause that matters, developing pattern recognition that later becomes the judgment senior professionals are actually paid for. Remove the reps, and the question of how the next cohort develops that judgment does not have an obvious answer, since AI can perform the task but cannot substitute for the person having done it themselves, repeatedly, under supervision, until the underlying pattern becomes second nature.

The broader reshaping of hiring patterns AI has driven across industries makes this dynamic more visible at scale, since it is not confined to any single profession. Law, consulting, and software engineering show the clearest data specifically because they have historically relied on large, structured entry-level cohorts performing well-defined, repeatable work, precisely the category of task current AI systems handle most capably.

A More Complicated Picture Than “AI Is Destroying Jobs”

The employment data does not point in a single, simple direction. Legal sector employment overall reportedly reached a ten-year high of approximately 1.24 million jobs in January 2026, even as the composition of that employment has shifted away from the traditional large junior-associate class toward roles built around AI-augmented workflows. That distinction matters: aggregate employment in a sector can hold steady or even grow while the specific entry point that used to train new professionals narrows considerably, a structural change that a single headline employment figure would not capture on its own.

Some employers appear to be responding by deliberately redesigning what an early-career role is meant to accomplish, rather than simply hiring fewer people into an unchanged structure. Reporting on professional services firms specifically describes efforts to rebuild graduate hiring pipelines and early-career training programs explicitly around AI-augmented work, treating the first year as an opportunity to teach judgment and oversight of AI output directly, rather than assuming that skill will develop naturally through the same volume of manual repetition it used to require.

What Rebuilding the Pipeline Might Actually Require

The broader question of how much human verification AI output genuinely requires turns out to be directly relevant to solving this problem, not just describing it. If junior professionals are no longer building expertise primarily by doing the work themselves, the alternative path likely runs through deliberately supervising and checking AI output early and often, a form of training that looks less like repetition and more like structured critical evaluation from the very first year rather than the fifth.

Whether that substitute genuinely builds the same depth of judgment that years of hands-on repetition once did is a question the professions most affected by this shift have not yet answered with any confidence, largely because not enough time has passed to observe how a cohort trained this way actually performs once they reach the seniority where that judgment is supposed to matter most. What the data reviewed here does make clear is that the shift is neither temporary nor confined to a single industry, and organizations that treat entry-level hiring reductions as a short-term cost saving, without also rebuilding a deliberate mechanism for developing the judgment that used to come from that work, may find the gap becomes considerably more expensive to close once today’s senior professionals begin to retire.

Share with Your Network

Join 231,000+ AI enthusiasts – Stay ahead with the latest insights and trends!

You may also like...