The Missing First Rung

In August 2025, researchers at Stanford University issued a warning: employment among 22- to 25-year-olds in the occupations most exposed to artificial intelligence had seen a relative decline of roughly 13%. Experienced workers were not the ones being sacked — it was the people trying to enter the labour market for the first time. Generative AI is not a threat looming in the future: it is already rewriting the world of work, and Generation Z is the first to experience it at first hand. This part examines which groups the shift affects most, and what lies ahead for new graduates.

1. Stanford's canaries

The August 2025 paper from researchers at Stanford's Digital Economy Lab has a title that speaks for itself: "Canaries in the Coal Mine?" The study examined US labour-market data from 2019 to 2025 and found a striking pattern. After ChatGPT arrived at the end of 2022, employment among workers aged 22 to 25 in the occupations most susceptible to generative AI saw a relative decline of roughly 13%. Relative is the operative word here: this is not necessarily a story of mass redundancies, but of young people simply being unable to get in — excluded from occupations that earlier generations entered as a matter of course.

One of the study's most important findings is that employment of more experienced workers — those over 30 — in those same occupations held steady. Generative AI — the form of artificial intelligence that can produce text, images and code — has not been making the people already in work redundant; it has been closing off the usual routes into these occupations. This is the entry-level bottleneck, and it matters because such jobs are precisely where young people are meant to gain the experience on which the rest of a career is built.

The canary metaphor is not incidental. Miners once carried canaries underground as living gas detectors: if the bird died, the men knew they were in danger. On the Stanford researchers' reading, 22- to 25-year-olds now play that role in the labour market: they are the first to show how far AI can displace human work. And the signs are not encouraging.

Researchers at Harvard Business School (Hosseini and Lichtinger, 2025) analysed the CVs of 65 million workers and data on more than 280,000 US firms, and found a similar pattern: at companies adopting generative AI, entry-level employment fell by roughly 9% six quarters after adoption, while employment of experienced staff showed no such break — indeed, it carried on rising. The fall was caused not by mass redundancies but by a drastic slowdown in hiring. The authors call this "seniority-biased" technological change: the tasks that disappear are exactly the routine ones usually given to people starting out.

None of this is a forecast. The shift is already under way, and it is being felt not by experienced specialists but by recent graduates, whose entry-level positions are disappearing.

2. Who is hit hardest? The occupations most exposed

According to an analysis published by the International Monetary Fund in January 2024, 40% of global employment is exposed to artificial intelligence, and in advanced economies that share reaches 60%. The risk is not spread evenly, however: AI is substantially reshaping some occupations, while others face no direct threat for the time being.

The Harvard research finds that financial services and technology are the two industries where entry-level job advertisements have fallen most sharply. Financial analysis, basic bookkeeping, writing standard reports, data processing — these are all tasks generative AI can now handle more efficiently and more cheaply than a new graduate can. In technology, junior software-development roles, quality-assurance testing and first-line IT support are thinning out; Amazon, Google and Microsoft have all said publicly that their AI coding assistants let them achieve the same output with fewer developers.

The sectors differ not only in their exposure, but in how quickly AI is replacing human labour. In copywriting, data entry and basic translation, generative AI can already do 70–80% of the work — and that share climbs month by month. In financial analysis and legal research, substitution is slower, because human judgement and human accountability remain inseparable from the work. In the skilled trades, healthcare and education, physical presence and human empathy are qualities AI cannot yet imitate — though robotics and remote-learning platforms are changing things there too. For the Stanford researchers, the decisive question is not whether an occupation "can be replaced", but how exposed its entry-level positions are to automation. And in almost every sector, the entry-level positions are the most exposed of all.

3. The gender divide — young men and young women face different risks

The labour-market changes AI has brought do not fall equally on young men and young women. The occupations more common among young women — marketing, communications, HR, customer service — are precisely the areas where generative AI can most readily substitute for human labour (as the story of Anna, the marketing professional, illustrated). The technical occupations more common among young men — programming, IT, engineering — are exposed too, but in those roles AI is more often used to augment workers than to replace them outright.

The differences extend well beyond the jobs men and women choose. In aggregated Gallup surveys from 2017 to 2024, 40% of women aged 18 to 29 describe themselves as liberal, against just 25% of men of the same age — a 15-percentage-point gap, the widest American generational research has ever recorded. That political polarisation is also reflected in the labour market: young men enter the technology sector and the skilled trades in greater numbers, while young women cluster in the service sector and the public sector — and the effects of automation are felt very unevenly across those sectors.

The widening divide is itself a social risk. Young men and young women now differ in political orientation, in their position in the labour market and in their mental-health outcomes — differences that together are creating a generational rift earlier cohorts did not experience. Where young people cannot find common ground — politically or economically — social cohesion weakens.

The labour-market differences affect mental health too. Among young men, unemployment and insecure work are more common, which is linked in turn to social withdrawal, gaming addiction and involvement in radical online communities. Among young women, on top of labour-market exclusion and unequal pay, the "double burden" of paid work and running a household weighs particularly heavily. The Harvard Youth Poll's 2025 figures put inflation (50%) and housing (41%) at the top of young women's list of concerns; young men's list is similar, but the priorities differ: a higher proportion of young men worry about job security, while women are more concerned about balancing a career with starting a family.

The gender gap surfaces in labour-market decisions as well. Young women are more likely to enter higher education — 60% of graduates in the United States are women — but a degree does not necessarily protect against AI-driven displacement. In the service sector, where women are more heavily represented, it is the communications, customer-service and content-production tasks that generative AI targets first. In the technology sector dominated by young men, meanwhile, the "junior gap" between experienced developers and new entrants keeps widening: firms prefer senior developers equipped with AI tools, who need less supervision. For both men and women, the underlying pattern is the same: entry-level opportunities are narrowing, and the gap between those who got in and those who were shut out keeps widening.

4. The gig-economy trap — when flexibility turns into insecurity

The narrowing of entry-level work reinforces something else: the spread of the gig economy, or platform economy. This is work carried out outside a conventional employment relationship, on a project-by-project basis and in short-term assignments — think of Uber drivers, food couriers, or the graphic designers who pick up jobs on Freelancer.com.

At first glance, the gig economy can look attractive to the young: flexible hours, quick money, no boss. The drawbacks are serious: no social insurance, no sick pay, no pension contributions, and no predictable income. For Generation Z — already stepping into an insecure labour market — the gig economy may be less a choice than a necessity: with no entry-level job on offer, platform work may be the only option left.

Deloitte's 2024 survey found that 56% of Generation Z live from one pay cheque to the next, and that 55% of them expect companies to address income inequality and access to jobs — markedly higher than among older generations. The spread of the gig economy deepens that inequality: the platforms generate the profits, while workers are left with unpredictable incomes and no social protection.

5. Consequences — what to watch

The displacement of entry-level work by AI is not a theoretical problem: it is already visible. These are the indicators worth watching:

The unemployment rate among new graduates. In Hungary, figures from Hungary's Central Statistical Office (KSH) show a headline unemployment rate of 13.9% among 15- to 24-year-olds in 2025 — more than three times the 4.4% rate for the population as a whole. The NEET rate — the share of young people not in employment, education or training — stood at 9.1% in that age group, which is concerning in its own right. The picture for new graduates is more nuanced still: graduate unemployment is lower, but among young people without a degree it is considerably higher, and among those working in the occupations AI threatens most it is expected to climb further.

The changing shape of job advertisements. The requirements in entry-level job postings are shifting too: in more and more of them, the ability to work alongside artificial intelligence is a basic requirement, while the number of roles focused entirely on routine tasks is falling. Employers, in other words, are already restructuring their expectations: anyone who cannot work with AI is at a disadvantage in the labour market.

The revival of the skilled trades. A curious counter-trend is visible as well: demand is rising for practical, physical trades — electricians, plumbers, joiners, chefs — because that work is still harder to automate. Deloitte's survey found that 59% of Generation Z are looking for work less vulnerable to automation, including skilled and manual trades. The trend may strengthen further. Yet the "blue-collar myth" — the stubborn conviction that physical, hands-on work is safe from automation — appears to be crumbling: the rapidly falling cost of humanoid robots and the spread of dark warehouses suggest that the trades are not a guaranteed refuge in the long run either.

In summary: Generative AI is not a threat looming in the future — it is already rewriting the world of work. The Stanford "canaries" research shows plainly that the biggest losers are not experienced workers but recent graduates, whose entry-level jobs are disappearing. The result is a vicious circle: with no entry-level work there is no experience; with no experience there is no advancement. And that is exactly the position Generation Z finds itself in — a generation the education system had already short-changed by comparison with their parents. The next part examines the economic and social burden this generation inherits — and how it is going to support an ageing society.

Young people are not losing their jobs — they are never offered them in the first place. As entry-level positions narrow, a generation is left with nowhere to learn how to work.

— The finding at the heart of the Stanford Digital Economy Lab's Canaries in the Coal Mine? (2025)

"Almost 40 percent of global employment is exposed to AI… In advanced economies, about 60 percent of jobs are exposed to AI, due to prevalence of cognitive-task-oriented jobs… In most scenarios, AI will likely worsen overall inequality."

— IMF Blog (2024)