Analysis and opinion
Picture a company celebrating its best year. Its experienced employees have AI assistants. Work moves faster. Customers get answers sooner. Nobody announces a round of redundancies.
But the graduate intake quietly disappears.
That imagined scenario captures an overlooked risk in the debate over AI and entry-level jobs. A business can look healthy while becoming much harder to enter. The damage would show up in an unanswered application, a cancelled traineeship or a junior vacancy that never gets advertised.
Our view: companies that use AI to remove the first rung of a career ladder should take responsibility for building another route up. The emerging evidence does not prove that AI is shutting out a generation. It gives us enough reason to stop treating training as somebody else’s problem.
The warning is about hiring, not a wave of sackings
In an update published on 12 August 2026, Stanford’s Digital Economy Lab reported a troubling pattern in US payroll records through June. Employment among 22–25-year-olds in occupations highly exposed to AI was about 19% below where it would have been if it had kept pace with less-exposed occupations for the same age group. Experienced workers showed no comparable gap. Stanford’s research update
That is a comparison between employment paths. It does not mean that AI fired 19% of young workers, or that 19% of graduates are unemployed.
The researchers found that the adjustment appeared mainly through weaker hiring. They also found no widespread, economy-wide job displacement associated with AI. Those two findings can coexist: the labour market can keep moving while one entrance becomes narrower. Stanford’s findings
The study is a working paper, and its authors explicitly describe the findings as patterns rather than proof of cause. Some differences began before the generative AI boom. Accounting for education reduces the gap, and the payroll sample shows a bigger divergence than national surveys. It is a warning signal, not a verdict. Full August 2026 paper
Experience has to come from somewhere
The uncomfortable possibility is that the easiest tasks to automate can also be the tasks through which newcomers learn.
Imagine a junior analyst preparing a rough research summary. The first version is imperfect. A colleague spots an unsupported conclusion, explains why a source is weak and asks for another draft. The document matters, but so does the correction.
If an experienced employee can produce that first draft with software, a manager may see an obvious saving. Whether that also removes a learning opportunity depends on how the work is redesigned.
The Stanford paper offers a related possible explanation: AI may substitute more readily for formal knowledge written down in procedures and textbooks, while complementing judgement acquired through experience. The authors present this as a mechanism to investigate, not a settled explanation. Stanford working paper
The risk is circular. Employers want experienced applicants. Applicants need employers willing to let them become experienced. Faster software does not automatically solve that problem.
The evidence also challenges the doom story
There is a serious counterargument: cheaper, faster work can help companies expand and hire more people.
A June 2026 working paper from researchers at Ramp and Revelio Labs linked AI spending with workforce records across 21,559 US firms. High-intensity AI adopters showed 10.2% higher employment over the first two years after adoption relative to the study’s comparison group. Entry-level employment was 12% higher. Low-intensity adopters showed no statistically detectable employment change. Ramp and Revelio paper
That does not prove AI created those jobs. Adopting firms were already different, including being larger and faster-growing. This is an observational study by company researchers, not a randomised experiment.
It also asks a different question from Stanford’s occupational analysis. A firm spending heavily on AI can grow while opportunities shift away from particular jobs elsewhere. Neither result, on its own, tells us what happens to every young applicant. Study design and limitations
So a blanket claim that AI is destroying entry-level employment goes beyond the evidence. A blanket assurance that newcomers have nothing to worry about does too.
AI could become the best coach a beginner ever had
There is another way to use these tools: give them to the people who are still learning.
A peer-reviewed study published in The Quarterly Journal of Economics in 2025 examined the introduction of an AI assistant among 5,172 customer-support agents. Productivity, measured as issues resolved per hour, improved by 15% on average. Less-experienced and lower-skilled workers improved both speed and quality. The researchers also found evidence of learning. Generative AI at Work
That finding matters because it challenges the idea that only experienced people can benefit. An assistant that surfaces useful knowledge could help a beginner become capable sooner.
The study concerns a specific workplace setting. It does not establish the effect on hiring across the economy, and its results cannot simply be transferred to every profession. But it demonstrates a credible alternative to treating AI as a reason to stop recruiting juniors.
A productivity win should include a way in
The choice for employers is practical. They can use a faster process to reduce recruitment, expand output, improve service or give newcomers more support. Different businesses will make different decisions.
Our argument is that companies should measure those decisions openly. Alongside hours saved, report the number of junior hires, access to mentoring and progression into more responsible work. If routine assignments disappear, replace them with supervised projects in which beginners explain, check and defend the result.
Keeping pointless busywork alive would be a poor answer. A real route to competence is the goal.
The strongest version of the AI boom is one in which more people can do valuable work. If its benefits flow mainly to people who already have experience, the technology may make companies more productive while making opportunity less accessible.
Join the discussion: Should businesses that automate junior tasks be expected to fund paid training routes—and would you trust AI to help train your first employee?
Leave a Reply