Will AI replace entry-level jobs?

AI is not so much replacing entry-level jobs as hollowing out the tasks inside them and raising the bar for what is left. The routine research, drafting and analysis that used to fill a junior's first years are exactly what the tools now do cheaply, so the openings shrink and the ones that remain expect judgement sooner. But the frightening number is not the graduate unemployment rate. The tasks being automated are the ones through which people used to build senior judgement. Remove them, and you can run a productive-looking operation for a few years while producing no senior people at all. The question worth asking is where the next generation of experts is supposed to come from once the rungs they used to climb have been automated away.

Three questions, not one#

"Will AI replace entry-level jobs?" is really three questions, and conflating them is what produces bad answers. Is AI replacing junior tasks? Yes, quickly, and that will continue. Is AI replacing junior jobs? Partly and unevenly: some hiring is slowing in the most exposed sectors, but many roles are being redefined rather than removed. And the third, which almost nobody asks: what happens to an organisation if the junior tasks disappear but the senior roles still require the experience those tasks used to create? That third question is where the real damage lives, and the one this page is about.

What the jobs data shows#

The clearest picture comes from PwC's Global AI Jobs Barometer, built from close to a billion job postings. It finds that entry-level roles most exposed to AI are now seven times more likely to require traditionally senior, human-intensive skills such as leadership and face-to-face judgement, and that these "seniorised" entry roles have grown by more than a third since 2019. In other words, the floor of the job is rising: the easy start is being automated, and what remains asks more of a newcomer than it used to. The World Economic Forum's 2025 Future of Jobs report names skills gaps as the single biggest barrier to transformation, and analytical thinking as the most valued skill, precisely the capability a hollowed-out junior role no longer builds by default.

The productivity research adds the twist. Brynjolfsson, Li and Raymond, studying 5,172 support agents, found AI raised the output of the newest and least experienced staff by thirty percent while barely moving the experts, because the tool hands expert patterns to novices. That is genuinely good for a graduate's first-week output. It is also the exact mechanism by which a junior can produce senior-looking work without doing the thinking that used to build the capability underneath it. The number goes up. The person does not.

The real risk: the missing rungs#

This is what I call the missing rungs problem: the junior tasks that used to carry people up to senior judgement are being removed by automation before anyone notices they were load-bearing. At the level of the individual it shows up as synthetic seniority, output that looks like ten years of judgement produced by someone who has not built it. Across an organisation it accumulates as capability debt, invisible while the outputs look fine, and expensive the moment a decision arrives that no junior has been developed to make. The saving from automating entry-level work is immediate and easy to book. The cost is deferred, compounding, and falls on the people who made the decision years later.

The same signal, outside the United States#

Most of the entry-level evidence in circulation is American, which makes it easy to dismiss as an artefact of one labour market. It is not. France's national statistics office, INSEE, reported in March 2026 that employment of 15 to 29 year olds, excluding apprentices, fell 7.4 percent year on year in IT services, 5.8 percent in publishing and 3.7 percent in management consulting in the fourth quarter of 2025, against minus 0.7 percent across the market sector as a whole. INSEE is explicit that this cannot be attributed to AI alone, and that caution should be carried with the number.

Korea's KDI found something sharper still. Where AI effects appeared, they fell not on the low-skilled but on younger, tertiary-educated workers and women. Germany's IAB, scoring more than nine thousand tasks across some 4,600 occupations, found substitutability rising around ten percentage points for degree-level expert occupations between 2019 and 2022 while remaining flat for helper occupations.

Three national research bodies, three methods, one direction: the pressure falls on educated entrants to judgement work. That inverts the assumption that automation comes for the least skilled first. It is the strongest available reason to treat the entry-level question as a capability problem rather than a wage problem.

Two honest caveats#

Two honest caveats. Labour-market data is noisy, and separating AI's effect on entry-level hiring from ordinary economic cycles is hard; some of the slowdown in graduate roles is macroeconomic, not machine. And the pipeline effect operates on a horizon of years, which means the strongest claims here, about senior talent shortages to come, are well-reasoned projections rather than measured outcomes. What is not in doubt is the direction: the tasks that built junior judgement are being automated, and organisations are mostly not redesigning how that judgement now gets built. That gap is the thing to act on before it is proven, because by the time it is proven the missing cohort is already missing.

Building new rungs instead of cutting the intake#

For organisations, the instinct to shrink graduate intake because the tasks are now cheap is the trap: it trades a visible saving for an invisible future liability. The better move is to redesign early-career development for a world where AI does the routine. Build new rungs on purpose to replace the ones automation removed, moments that develop judgement directly rather than as a by-product of grunt work. Keep some work deliberately unaided so juniors still practise the reasoning the tools would otherwise do for them. Redesign apprenticeship and graduate programmes around capability, not task completion. And measure whether people are becoming capable, not just whether the output is good, because output has stopped being a reliable signal. For individuals starting out, the same logic points to building the skills that survive AI, judgement, framing and the willingness to do the hard thinking yourself, because those are now the differentiator that the automated tasks used to hide.

25 September 2026: the apprenticeship subsidy, and the evaluation bottleneck#

Two documents in one week put a mechanism under the pattern this page describes. Patrick Harker, a Wharton professor and until 2025 president of the Federal Reserve Bank of Philadelphia, argued in Fortune on 25 September, under the headline that entry-level jobs “were actually secret apprenticeships all along”, that the junior’s routine output offset the cost of training the next generation of seniors, and that once a machine does the routine the offset goes and the case for the hire with it: “Firms are not firing their junior employees; they’re hiring fewer of them.” He cites the Stanford Digital Economy Lab’s finding that workers aged 22 to 25 in the most AI-exposed occupations are running roughly 19 per cent behind peers in less-exposed fields, and records both the authors’ own caveat that these are “descriptive patterns, not causal estimates” and a rival analysis by Iscenko and Curto Millet of 238 million job postings that attributes the decline to interest rates rather than to AI; the figure is under dispute and is reported here as a claim. His prescription matches the section above: stop treating junior hiring as a cost line and fund training, rotations, mentoring and coaching as capital investment. The second document is a model posted on 24 September by Itai Ashlagi, Ramesh Johari, Jon Kleinberg and Anushka Murthy on what AI-written applications do to hiring: the documents stop telling candidates apart, firms fall back on prior experience, and the inexperienced but well-matched lose most. Its predicted response, an intermediate assessment that produces evidence a CV cannot, is examined on should AI reject an application before a human reads it. Personnel Today (Maddison Frost, 24 September) reported Klarus research in which 19 per cent of employers said they had reduced recruitment into early-career roles and 42 per cent of those attributed it to AI, and quoted Cheney Hamilton, a director at Bloor: “AI raises the floor of what a junior can produce, but it doesn’t give them the judgement to know when the output is wrong, and that judgement historically came from repetition and consequence.”

Key research and primary sources

• INSEE, France (2026). Note de conjoncture: digital investment, artificial intelligence and youth employment. Institut national de la statistique et des etudes economiques, March 2026, published in French. graded entry.
• Korea Development Institute (KDI) (2023). Changes in the labour market due to artificial intelligence and policy directions (Research Report 2023-03). KDI, published in Korean. graded entry.
• Institut fur Arbeitsmarkt- und Berufsforschung (IAB), Germany (2024). Folgen des technologischen Wandels fur den Arbeitsmarkt (Consequences of technological change for the labour market: it is above all the highly qualified who feel digitalisation). IAB-Kurzbericht 5/2024, published in German. graded entry.
• Pearson and AWS (2026). AI Readiness: Building the Bridge from Higher Education to Work. Pearson and Amazon Web Services, 13 April 2026.
• PwC (2026). 2026 Global AI Jobs Barometer: Two futures for jobs in an AI era. PwC, 15 June 2026.
• PwC (2025). Global AI Jobs Barometer, on seniorised entry-level roles.
• World Economic Forum (2025). The Future of Jobs Report 2025.
• Brynjolfsson, E., Li, D. and Raymond, L. (2025). Generative AI at Work. Quarterly Journal of Economics, 140(2), 889-942. Earlier version NBER Working Paper 31161.

Related SuperSkills research#

This connects to the missing rungs, synthetic seniority, the missed reps, capability debt and human skills in the age of AI. On how juniors build capability when AI does the practice, see how juniors become senior and how humans learn with AI, and on the individual career response, staying valuable in the age of AI. The individual version of the question is will AI replace my job. For how the discourse itself changed between 2023 and 2026, and why the founding estimates are still quoted over the measurements that complicate them, see the best writing on AI. See should I still learn to code. See which jobs are safest from AI.

About this research#

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. This work draws on research across more than 200 organisations in 30 countries over seven years. Findings are attributed to the studies that produced them and kept separate from the interpretation and named concepts, which are the author's. This is a living reference, reviewed and updated as significant new evidence appears.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.