Francesco Di Costanzo
Back to articles

(16) AI Is Quietly Killing the Entry-Level Job

/15 min read

The Hiring Tap Is Turning Off

The numbers arrived before the narrative did. In November 2025, a Stanford team led by Erik Brynjolfsson published "Canaries in the Coal Mine," drawing on ADP administrative payroll data covering millions of American workers. The headline finding: a 16 percent relative decline in employment for 22- to 25-year-olds in occupations most exposed to artificial intelligence, concentrated in the period since late 2022. Software developers in that age bracket saw nearly a 20 percent drop from their pre-ChatGPT peak. The decline was sharpest in occupations where AI substitutes for human work rather than augmenting it.

A separate Harvard study, published around the same time and drawing on a different dataset — 62 million workers across 285,000 firms, sourced from LinkedIn and Revelio Labs — found a converging pattern. Junior-level employment fell approximately 8 percent at firms that adopted AI tools, visible within six quarters of ChatGPT's launch. The mechanism was not mass layoffs. Separation rates actually declined slightly. Firms were hiring 5.0 fewer junior workers per quarter. They were turning down the tap, not pushing people out.

A King's College London study documented the same dynamic in the United Kingdom. Firms with high AI exposure reduced total employment by 4.5 percent on average, with junior positions falling 5.8 percent. Highly exposed firms were 16.3 percentage points less likely to post new vacancies. Technical roles — software engineers, data analysts — saw the steepest declines, and high-paying firms experienced a 9.6 percent employment drop compared to near zero for lower-paying firms.

Survey data fills in the picture. Resume.org polled 933 U.S. business leaders in February 2026 and found that 21 percent of companies had already frozen entry-level hiring due to AI, with 47 percent expecting to eliminate entry-level hiring entirely by 2027. Randstad and the World Economic Forum, analysing 126 million job postings worldwide, reported that entry-level postings had fallen 29 percentage points since January 2024.

Bank of America's research institute flagged a related milestone: the unemployment rate for recent college graduates aged 22 to 27 now exceeds the overall workforce rate for the first time since before the pandemic. Thirteen percent of unemployed Americans were "new entrants" seeking their first job — the highest share since 1988. The broad pattern is consistent. Entry-level white-collar hiring has measurably declined, and the mechanism is not displacement from existing roles but a sustained reduction in the rate at which firms bring new workers in.

Pre-emptive Cuts for a Technology That Hasn't Arrived

The most striking feature of the data is the timing. The Harvard study found the decline in junior hiring visible within a single quarter of ChatGPT's launch in November 2022. That is far too fast for firms to have deployed AI tools at scale, measured their performance, and restructured headcount accordingly. What happened instead is that firms anticipated automation and adjusted hiring before proving it worked.

Harvard Business Review confirmed this interpretation in January 2026. In a survey of organisations making AI-related workforce changes, only 2 percent had made large reductions based on proven AI performance. Twenty-nine percent were hiring fewer workers in anticipation of what AI might eventually deliver. Companies were acting on expectation, not results. The gap between AI rhetoric and AI deployment remains wide: a KPMG CEO Outlook pulse survey from March 2026 found that fewer than 10 percent of CEOs attributed planned workforce reductions to AI, and 67 percent of executives said they had only an "initial perspective" on how AI would affect their workforce. Yet 55 percent expected AI to eventually increase hiring — a reminder that executive sentiment is deeply heterogeneous and internally contradictory.

Goldman Sachs projected that AI-related displacement could raise the U.S. unemployment rate to 4.5 percent by the end of 2026. Anthropic's own labour market research noted no systematic increase in unemployment but found "suggestive evidence" that hiring of younger workers has slowed, with occupations facing higher AI exposure projected to grow less through 2034.

The pattern is familiar: firms act on the option value of a technology before its payoff is certain. The risk is that anticipatory cuts, if widespread enough, create the labour market weakness they were designed to hedge against.

When You Automate the Job, You Automate the Learning

Entry-level roles have always served a dual purpose: production and professional development. A first-year analyst does not just produce financial models; the analyst learns to produce financial models. A junior associate does not just draft memos; the associate builds legal reasoning through the act of drafting. Wharton's Cornelia Jessie Roventa has described this as "cognitive apprenticeship" — the process by which novices acquire expert judgment through structured repetition of progressively complex tasks. When AI automates those tasks, it eliminates both the output and the training mechanism that produced it. Roventa identifies a "delayed feedback problem": the consequences of cutting junior training do not surface until several years later, when firms discover they lack the mid-level talent they need.

The task-level evidence supports this concern. Revelio Labs tracks the share of AI-exposed tasks appearing in job postings. That share declined from 29 percent in early 2022 to 25.5 percent by early 2025. The tasks disappearing are specific: financial transaction processing, insurance and tax paperwork, reconciliations, basic code debugging — precisely the routine work that constituted the training ground for junior professionals in finance, law, and technology.

The legal sector offers the most detailed case study. AI adoption among legal professionals doubled in a single year, rising from 31 percent in 2025 to 69 percent in 2026, according to the 8am Report. Thirty-nine percent of legal professionals predict a reduction in paralegal and support roles. Twenty-one percent anticipate fewer junior associate positions. Reuters and Citi-Hildebrandt reported that 86 percent of large law firms plan to grow their associate ranks through 2027, but only 35 percent plan to increase first-year lawyers and 37 percent plan to increase summer associates. The law firm pyramid — wide at the base, narrow at the top — is compressing into a cylinder. Sixty-three percent of firms expect generative AI to modify leverage models by 2035. Clifford Chance cut approximately 50 London back-office roles in late 2025, citing AI alongside offshoring and reduced demand.

The accounting and consulting professions show parallel shifts. PwC announced plans to cut U.S. campus hiring by nearly a third, with audit associates projected to drop 39 percent by 2028. PwC's UK arm hired 200 fewer entry-level workers in 2025 than in 2024. The Big Four are deploying AI at a scale that makes the shift structural, not experimental. EY launched EY.ai, giving 80,000 tax staff access to 150 AI agents, with plans to scale to 100,000 agents by 2028. Deloitte partnered with Anthropic to deploy Claude to 470,000 employees and rolled out Zora AI, an agentic system developed with Nvidia. KPMG stated directly: "We want juniors to become managers of agents." Consulting firms have frozen entry-level salaries for a third consecutive year, with AI enabling a shift from the traditional pyramid staffing model toward what some describe as an "obelisk" or "hourglass."

Klarna offers a compressed version of the same story. The fintech halved its workforce from approximately 5,500 to 3,000 through a hiring freeze and attrition — not layoffs — while promoting an AI chatbot it said was doing the work of 700 to 800 customer service agents. The company later partially reversed its AI-first strategy after quality complaints, illustrating the gap between AI capability in controlled settings and AI performance at production scale.

Maybe This Is Just a Recession With Better Branding

The counterargument deserves honest engagement. Not all of the entry-level decline is attributable to AI. Indeed's Hiring Lab found that junior postings fell approximately 7 percent year-on-year in September 2025, broadly in line with overall market declines — suggesting that some of what looks like an AI effect is simply cyclical weakness in the labour market. LinkedIn's monthly data showed U.S. hiring running 16 percent below pre-pandemic levels, with a 5.7 percent year-on-year decline in January 2026. The labour market is soft for everyone, not only for entry-level workers.

Brookings published a direct challenge in March 2026, finding that the pace of occupational change since ChatGPT's release is comparable to what followed the personal computer in 1984 and the commercial internet in 1996 — periods that ultimately generated net employment growth. An NBER working paper using Danish administrative data found null effects on earnings and hours within two years of AI adoption. Yale's Budget Lab detected no economy-wide break in occupational mix attributable to generative AI. The Economic Innovation Group critiqued the Stanford "Canaries" paper specifically, arguing that the narrow 22-to-25 age band is mechanically sensitive to hiring inflows — if hiring slows for any reason, this cohort shrinks through aging even without AI-specific displacement.

At the international level, the IMF estimated that 40 percent of global jobs are exposed to AI-driven change but noted that AI also creates new categories of work. The World Economic Forum projected a net gain of 78 million jobs by 2030. OECD surveys found that workers are generally more positive about AI's impact on their own jobs than public discourse would suggest. LinkedIn's own data contained a notable counterexample: the entry-level share in customer support actually increased by 1.1 percentage points from 2022 to 2025, even as the sector is widely considered among the most AI-exposed.

The most honest reading of the evidence is that a structural wedge is forming on top of cyclical weakness. The Stanford and Harvard studies control for firm-level and economy-wide shocks, isolating an additional AI-specific effect. Both factors are operating simultaneously. The question is not whether AI or the macro cycle is responsible — it is how much of the entry-level decline will reverse when the cycle turns, and how much is permanent.

AI Could Accelerate Training — If Firms Still Hire the Trainees

The strongest counterargument to the apprenticeship-collapse thesis is that AI itself can compress learning curves. A peer-reviewed field study published in the Quarterly Journal of Economics examined a Fortune 500 customer-support operation and found that AI raised worker productivity by approximately 15 percent, with the gains disproportionately benefiting less experienced workers. Employees with two months of tenure using AI performed at the level of employees with six months of tenure working without it. A separate experimental study by Noy and Zhang, published in Science, found that access to ChatGPT decreased task completion time by 40 percent and raised output quality by 18 percent, with less-skilled workers capturing the largest gains.

The implications cut both ways. If AI compresses the time it takes a novice to reach competence, the training pipeline does not necessarily break — it accelerates. PwC's AI assurance leader, Jenn Kosar, has argued that AI-equipped new hires will reach manager-level capability within three years rather than the traditional five to seven. Forbes described the emerging "Junior Accountant 2.0" model in which entry-level workers shift from processing to oversight, from reactive execution to analytical judgment. A Deloitte survey of 11,000 workers across 17 countries found that 61 percent believe AI can support upskilling for entry-level employees.

The paradox is that AI compresses learning curves only if firms continue to hire learners. The augmentation story requires a junior worker in the seat, using AI tools, building judgment through supervised practice. If firms eliminate the seat entirely — treating AI as a replacement for the worker rather than a tool the worker uses — the learning-acceleration benefit never materialises. The distinction between redesigning entry-level roles and eliminating them is the dividing line between a future in which junior professionals develop faster and one in which they do not develop at all.

A Margin Gain Now, a Talent Crisis Later

The short-term economics are straightforward. Fewer junior salaries, faster output per remaining worker, immediate margin improvement. McKinsey estimates that current technologies could automate approximately 57 percent of U.S. work hours in theory — but OECD data shows that only 20 percent of firms were actually using AI in 2025, up from 8.7 percent in 2023. Adoption is accelerating but remains early. The long-term question is different: if firms stop investing in junior talent, who becomes the next generation of senior analysts, partners, and managing directors?

The offshoring parallel is instructive. In the 1990s and 2000s, companies moved junior analytical, administrative, and IT roles to lower-cost geographies. A 2005 Deloitte survey found that a quarter of companies that outsourced eventually reversed their decisions, citing hidden costs and quality erosion. The "poaching externality" compounds the problem: firms that invest in training juniors lose those workers to competitors that did not invest, creating a collective-action problem that discourages training industry-wide. Two-thirds of hiring managers already say that entry-level hires arrive underprepared, according to Deloitte — a gap that will widen if the volume of on-the-job training declines.

No empirical study has yet measured the downstream effect of reduced junior hiring on senior talent supply. The timeline is too short. But the direction of the evidence is clear enough to warrant attention. The firms that recognise entry-level roles as an investment in future capacity — and redesign those roles around AI augmentation rather than eliminating them — will compound an advantage that becomes visible only when their competitors discover the gap.

Sources

Employment Data and AI-Labour Market Studies

  1. Stanford Digital Economy Lab, "Canaries in the Coal Mine: Six Facts About the Recent Employment Effects of Artificial Intelligence" https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/

  2. Observer, "AI Is Shrinking the Job Market for Junior Workers, Harvard Study Finds" https://observer.com/2025/09/ai-shrinking-job-market-junior-workers-harvard-study/

  3. Brynjolfsson, Li, and Raymond, "Generative AI at Work" https://academic.oup.com/qje/article/140/2/889/7990658

  4. Resume.org, "1 in 5 Companies Have Stopped Hiring Entry-Level Workers Because of AI" https://www.prnewswire.com/news-releases/resumeorg-survey-1-in-5-companies-have-stopped-hiring-entry-level-workers-because-of-ai-302707385.html

  5. Revelio Labs, "The Tasks You Won't See in Job Postings Anymore" https://www.reveliolabs.com/news/macro/the-tasks-you-won-t-see-in-job-postings-anymore/

  6. Revelio Labs, "Is AI Responsible for the Rise in Entry-Level Unemployment?" https://www.reveliolabs.com/news/macro/is-ai-responsible-for-the-rise-in-entry-level-unemployment/

  7. King's College London, "New Study Reveals Early Impact of AI on Job Market in UK" https://www.kcl.ac.uk/news/new-study-reveals-early-impact-of-ai-on-job-market-in-uk

  8. Fortune, "First-of-Its-Kind Stanford Study on AI and Entry-Level Jobs" https://fortune.com/2025/08/26/stanford-ai-entry-level-jobs-gen-z-erik-brynjolfsson/

  9. Fortune, "Gen Z Unemployment Rate for Recent College Graduates — Bank of America" https://fortune.com/2025/09/04/gen-z-unemployment-rate-recent-college-gradutes-bank-of-america/

  10. LinkedIn, "Labor Market Report: Building a Future of Work That Works" https://economicgraph.linkedin.com/content/dam/me/economicgraph/en-us/PDF/linkedIn-labor-market-report-building-a-future-of-work-that-works-jan-2026.pdf

  11. Adzuna UK, "Job Market Report — 30 June 2025" https://www.adzuna.co.uk/job-market-report/30-june-2025-2/

  12. LinkedIn, "US Monthly Insights — February 2026" https://economicgraph.linkedin.com/content/dam/me/economicgraph/en-us/PDF/US%20Monthly%20Insights%20%E2%80%93%20Feb%202026.pdf

  13. Anthropic, "Labor Market Impacts" https://www.anthropic.com/research/labor-market-impacts

Anticipatory Behaviour and Executive Sentiment

  1. Harvard Business Review, "Companies Are Laying Off Workers Because of AI's Potential — Not Its Performance" https://hbr.org/2026/01/companies-are-laying-off-workers-because-of-ais-potential-not-its-performance

  2. KPMG, "CEO Outlook Pulse 2026" https://kpmg.com/us/en/media/news/ceo-outlook-pulse-2026.html

  3. Goldman Sachs, "How Will AI Affect the Global Workforce?" https://www.goldmansachs.com/insights/articles/how-will-ai-affect-the-global-workforce

  4. Yahoo Finance, "Goldman Sachs Warns AI-Fueled Layoffs Could Raise the Unemployment Rate" https://finance.yahoo.com/news/goldman-sachs-warns-ai-fueled-layoffs-could-raise-the-unemployment-rate-this-year-chart-154251740.html

Counterarguments and Macro-Cyclical Evidence

  1. Indeed Hiring Lab, "September Labor Market Squeeze on New Entrants" https://www.hiringlab.org/2025/09/25/september-labor-market-squeeze-on-new-entrants/

  2. Brookings, "Research on AI and the Labor Market Is Still in the First Inning" https://www.brookings.edu/articles/research-on-ai-and-the-labor-market-is-still-in-the-first-inning/

  3. NBER Working Paper 33777, "AI Adoption and Labour Market Effects in Denmark" https://www.nber.org/papers/w33777

  4. Yale Budget Lab, "Evaluating the Impact of AI on the Labor Market" https://budgetlab.yale.edu/research/evaluating-impact-ai-labor-market-novemberdecember-cps-update

  5. Economic Innovation Group, "Critique of Canaries in the Coal Mine" https://eig.org/wp-content/uploads/2026/01/TAWP-Iscenko-Millet.pdf

  6. IMF, "New Skills and AI Are Reshaping the Future of Work" https://www.imf.org/en/blogs/articles/2026/01/14/new-skills-and-ai-are-reshaping-the-future-of-work

  7. IMF Staff Discussion Note, "Bridging Skill Gaps for the Future" https://www.imf.org/-/media/files/publications/sdn/2026/english/sdnea2026001.pdf

  8. WEF, "The Future of Jobs Report 2025" https://www.weforum.org/publications/the-future-of-jobs-report-2025/

  9. Law and Economics Center, "AI, Productivity, and Labor Markets: A Review of the Empirical Evidence" https://laweconcenter.org/resources/ai-productivity-and-labor-markets-a-review-of-the-empirical-evidence/

  10. OECD, "AI Use by Individuals Surges Across the OECD as Adoption by Firms Continues to Expand" https://www.oecd.org/en/about/news/announcements/2026/01/ai-use-by-individuals-surges-across-the-oecd-as-adoption-by-firms-continues-to-expand.html

Productivity, Training, and Augmentation

  1. Noy and Zhang, "Experimental Evidence on the Productivity Effects of Generative AI" https://www.science.org/doi/10.1126/science.adh2586

  2. PubsOnLine/Management Science, "The Effects of Generative AI on High-Skilled Work" https://pubsonline.informs.org/doi/10.1287/mnsc.2025.00535

  3. Wharton, "Is AI Pushing Us to Break the Talent Pipeline?" https://knowledge.wharton.upenn.edu/article/is-ai-pushing-us-to-break-the-talent-pipeline/

  4. Harvard Business Review, "The Perils of Using AI to Replace Entry-Level Jobs" https://hbr.org/2025/09/the-perils-of-using-ai-to-replace-entry-level-jobs

  5. Forbes Tech Council, "Rethinking the Apprenticeship Ladder in an AI-Powered Workplace" https://www.forbes.com/councils/forbestechcouncil/2026/02/23/rethinking-the-apprenticeship-ladder-in-an-ai-powered-workplace/

  6. Forbes Tech Council, "Junior Accountant 2.0: How AI Agent Adoption Rewrites Entry-Level Career Paths" https://www.forbes.com/councils/forbestechcouncil/2026/02/19/junior-accountant-20-how-ai-agent-adoption-rewrites-entry-level-career-paths/

  7. Deloitte, "AI, Demographic Shifts, and Agility" https://www.deloitte.com/us/en/insights/topics/talent/strategies-for-workforce-evolution.html

  8. Deloitte, "Gen Z and Millennial Survey 2025" https://www.deloitte.com/global/en/issues/work/genz-millennial-survey.html

Sector-Specific Evidence: Legal

  1. Reuters/Citi-Hildebrandt, "Is Big Law's Pyramid Due an AI Makeover?" https://www.reuters.com/legal/government/is-big-laws-pyramid-due-an-ai-makeover-2025-12-11/

  2. LawNext/8am Report, "AI Adoption Among Legal Professionals Has More Than Doubled in a Year" https://www.lawnext.com/2026/03/ai-adoption-among-legal-professionals-has-more-than-doubled-in-a-year-new-8am-report-finds-but-firms-lag-far-behind-individual-practitioners.html

  3. Legal Cheek, "Clifford Chance to Cut 50 London Support Roles" https://www.legalcheek.com/2025/11/clifford-chance-to-cut-50-london-support-roles/

  4. Heather Suttie, "The Law Firm Pyramid Rollover" https://heathersuttie.ca/insights/the-law-firm-pyramid-rollover/

Sector-Specific Evidence: Accounting, Consulting, and Technology

  1. Business Insider, "PwC Hiring Fewer Junior Associates — AI, Offshoring, Big Four" https://www.businessinsider.com/pwc-hiring-fewer-junior-associates-ai-offshoring-big-four-2025-8

  2. Business Insider, "PwC UK Cuts Graduate-Level Hiring" https://www.businessinsider.com/pwc-uk-cuts-graduate-level-hiring-ai-offshoring-2025-9

  3. Business Insider, "4 Key Ways AI Changed the Big Four in 2025" https://www.businessinsider.com/how-ai-changed-big-four-workflow-hiring-jobs-2025-12

  4. LinkedIn/McKinsey, "Consulting Firms Use AI to Squeeze Entry-Level Hires" https://www.linkedin.com/news/story/consulting-firms-use-ai-to-squeeze-entry-level-hires-6781876/

  5. Fortune, "Klarna CEO Says Tech CEOs Sugarcoating AI Impact on Jobs" https://fortune.com/2025/10/10/klarna-ceo-sebastian-siemiatkowski-halved-workforce-says-tech-ceos-sugarcoating-ai-impact-on-jobs-mass-unemployment-warning/

  6. CNBC, "Klarna CEO Says AI Helped Company Shrink Workforce by 40%" https://www.cnbc.com/2025/05/14/klarna-ceo-says-ai-helped-company-shrink-workforce-by-40percent.html

  7. Fast Company, "Klarna Tried to Replace Its Workforce With AI" https://www.fastcompany.com/91468582/klarna-tried-to-replace-its-workforce-with-ai

Workforce Projections and Adoption Data

  1. BLS, "AI Impacts in BLS Employment Projections" https://www.bls.gov/opub/ted/2025/ai-impacts-in-bls-employment-projections.htm

  2. McKinsey, "Superagency in the Workplace" https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/superagency-in-the-workplace-empowering-people-to-unlock-ais-full-potential-at-work

  3. Randstad/WEF, "Gen Z and the Competitive Job Market" https://www.weforum.org/stories/2025/09/gen-z-are-competitive-job-market-randstad/

  4. WEF, "Top Jobs and Labour Market Stories 2025" https://www.weforum.org/stories/2026/01/top-jobs-and-labour-market-stories-2025/

  5. Deloitte, "AI Adoption Challenges and AI Trends — Quarterly Pulse" https://www.deloitte.com/us/en/what-we-do/capabilities/applied-artificial-intelligence/blogs/pulse-check-series-latest-ai-developments/ai-adoption-challenges-ai-trends.html

Talent Pipeline and Offshoring Parallels

  1. IT Pro, "Are We Facing an AI-Fueled Talent Pipeline Time Bomb?" https://www.itpro.com/business/careers-and-training/are-we-facing-an-ai-fueled-talent-pipeline-time-bomb

  2. Sound of Development, "The History of Offshoring: Rise, Fall, and Lessons for AI" https://soundofdevelopment.substack.com/p/the-history-of-offshoring-rise-fall