Start with one number. At the same companies, headcount for developers aged 22 to 25 has fallen by almost 20 percent since 2024, while headcount for developers over 30 grew. That is not a story about artificial intelligence getting cleverer. That is a story about who a company still thinks is worth training.
For this piece I pulled fourteen sources: Stanford, the UK Institute of Student Employers, the British Standards Institution, Randstad, Gartner, Boston Consulting Group, Anthropic, PwC, Fabric, PNAS, and a few more. Read that again if you need to. Fourteen separate studies, and they do not all agree on why the entry-level job is disappearing. That disagreement is the actual story.
The First Rung Is Being Sawn Off
The UK Institute of Student Employers tracks graduate recruitment every year. Their early-careers benchmark work points to tech graduate roles down 46 percent from 2024, with a further 53 percent drop projected. That is not a slowdown. That is a floor giving way.
The British Standards Institution ran a seven-country survey and found that 82 percent of business leaders are cutting or actively re-evaluating entry-level roles because of automation. Eighty-two percent. Nearly every leader in the room has already decided the graduate hire is the first line item to question.
Randstad looked at roughly 126 million job postings worldwide. Roles asking for zero to two years of experience are down 29 percentage points since January 2024. That is the plainest evidence there is. Companies are not quietly reducing junior hiring. They are rewriting the job description so the junior candidate never applies.
Everyone Blames AI. Less Than One Percent Of Layoffs Actually Prove It.
Here is where it gets uncomfortable for the people doing the cutting. Gartner has been tracking what it calls AI-washing, where companies attribute layoffs to artificial intelligence in the press release, without the productivity data to back it up. Their finding: fewer than 1 percent of layoffs are tied to a measurable AI productivity gain.
So the graduate job is disappearing. That part is real. But the reason given in the boardroom memo is, in the vast majority of cases, not the reason on the spreadsheet. AI is a convenient story. Restructuring, cost control, and margin pressure are the less flattering ones underneath it.
Boston Consulting Group offers a cleaner way to think about which jobs are genuinely at risk versus which are just being used as cover. Their framework plots roles on two axes: how much the job depends on human interaction and judgement, and how repeatable and structured the underlying process is. High repeatability, low judgement, that role gets substituted. High judgement, low repeatability, that role gets augmented, not replaced. Most entry-level roles sit closer to the first category than companies like to admit when they are also trying to attract graduate applicants for the roles that remain.
What Is Actually Being Automated
Anthropic built something called the Observed Exposure framework, which combines what a model can theoretically do with what people are actually asking it to do across real API and chat usage. That distinction matters. Theoretical capability and observed use are not the same thing, and the gap between them is where a lot of the panic and a lot of the denial both live.
PwC’s Global AI Jobs Barometer, built from over a billion job ads, found that the skills required in AI-exposed roles are changing 66 percent faster than in roles with low exposure. The job has not vanished so much as it has been rewritten underneath the person doing it, faster than most training pipelines can keep up.
The People Coming Up Are Learning To Perform AI Use, Not Master It
This is the part that should worry any hiring manager more than the headline layoff numbers. Fabric analysed roughly 19,400 technical interviews and found AI-assisted cheating flagged in interviews rising from 9 percent to 38.5 percent. Candidates are not learning to use these tools well. They are learning to use them invisibly.
A study published in the Proceedings of the National Academy of Sciences tested whether people can actually tell AI-generated text from human-written text. Across 4,600 participants, accuracy sat at 50 to 52 percent. That is coin-flip territory. The heuristics people rely on to spot machine writing are wrong more often than they are right.
Anthropic’s own research, drawn from around 10,000 conversations, maps something they call the 4D AI Fluency Framework across eleven observable behaviours. The pattern that comes out of it: iteration drives fluency, people get better at using the tool the more they go back and forth with it, but discernment and fact-checking lag well behind. People are becoming fast. They are not becoming careful.
A study in Frontiers in Education looked at postgraduate researchers and found something similar from the other direction: high ethical awareness about when AI use is appropriate, paired with lagging technical understanding of what the tool is actually doing under the surface. Good instincts, thin foundations.
The Credential Rush
None of this has stopped people from trying to skill up. Skillsoft reported a 994 percent year-on-year increase in completions of AI-related skill benchmark assessments on its Percipio platform. LinkedIn’s broader Work Trend Index and Learning data point the same direction, a sharp rise in AI-foundation credentialing activity, even if the platform does not publish a single clean dataset naming the exact volume share.
So here is where all fourteen studies land in the same place, even when they disagree on the mechanism. The entry-level rung is being removed faster than anyone is being taught to climb without it. Companies are citing AI for cuts that mostly are not about AI productivity at all. And the graduates racing to catch up are getting faster at using the tools and no better at knowing when to trust them.
That last part is the one worth sitting with. Speed without discernment is not a skill. It is just a faster way to be wrong.
Want more? Change your life. Subscribe to MONDAY INFLUENCER®
That last part is the one worth sitting with. Speed without discernment is not a skill. It is just a faster way to be wrong.
Want more? Change your life. Subscribe to MONDAY INFLUENCER®
References
Entry-Level Hiring Data
Stanford Institute for Human-Centered Artificial Intelligence (HAI), 2026 AI Index Report — https://hai.stanford.edu/ai-index
Institute of Student Employers (ISE), Graduate Recruitment Research Portal — https://ise.org.uk/knowledge/research/
British Standards Institution (BSI), Evolving Together: Flourishing in the AI Workforce — https://www.bsigroup.com/en-US/insights-and-media/media-center/press-releases/2024/september/embrace-ai-tools/
Randstad, The Gen Z Workplace Blueprint: Future Focused, Fast Moving — https://www.randstad.com/press/2025/genz-workplace-blueprint/
AI-Washing And Layoffs
Gartner, Layoff Messaging in the Age of AI — https://www.gartner.com/en/articles/lay-off-messaging
Boston Consulting Group, AI Will Reshape More Jobs Than It Replaces — https://www.bcg.com/publications/2026/ai-will-reshape-more-jobs-than-it-replaces
What Is Actually Being Automated
Anthropic Research, Labor market impacts of AI: A new measure and early evidence — https://www.anthropic.com/research/labor-market-impacts
PwC Global, 2026 AI Jobs Barometer — https://www.pwc.com/id/en/services/reimagine-digital/ai-jobs-barometer.html
Human Judgement And AI Literacy
Fabric, State of AI Interview Cheating in 2026 — https://www.fabrichq.ai/
Jakesch et al., Human heuristics for AI-generated language are flawed, Proceedings of the National Academy of Sciences, 2023 — https://www.pnas.org/doi/10.1073/pnas.2208839120
Anthropic Research, AI Fluency Index — https://www.anthropic.com/research
Jin et al., AI literacy as contextual judgement: postgraduate researchers’ appropriateness judgements of generative AI use in a multilingual Chinese setting, Frontiers in Education, 2026 — https://www.frontiersin.org/journals/education/articles/10.3389/feduc.2026.1886090/full
The Upskilling Response
Microsoft and LinkedIn, Work Trend Index / LinkedIn Learning AI Upskilling Framework — https://www.linkedin.com/business/talent/blog/learning-and-development/a-new-framework-for-ai-upskilling-across-your-organisation
Skillsoft, AI Skill Benchmark Completion Data — https://press.skillsoft.com/