First: “exposed to AI” does not mean “your job disappears”
That distinction matters because many scary headlines quietly mix up two different questions: Can AI perform some of the tasks inside a job? and Will the entire job stop existing?
The International Labour Organization’s 2025 global index found that about one in four workers are in occupations with some degree of generative-AI exposure. But only 3.3% of global employment fell into its highest exposure category. The ILO’s conclusion was not that one quarter of humanity is about to become unemployed. It said that, because most occupations still require human input, transformation is more likely than total redundancy for many jobs.
In 2026, the ILO issued another warning that is easy to miss: exposure indicators are useful for identifying tasks that could be affected, but they should not be read as direct predictions of job losses. A job can be highly exposed and still survive — with different tools, fewer routine tasks, higher output expectations or a different mix of workers.
Some jobs really will disappear — while other jobs grow
The World Economic Forum’s Future of Jobs Report 2025, based on responses from more than 1,000 employers, estimated that broad labour-market changes could create 170 million roles and displace 92 million by 2030, for a net increase of 78 million. Those numbers are employer projections, not a guarantee, and they include forces beyond AI such as demographics, economic change and the green transition. Still, they show why “technology destroys jobs” and “technology creates jobs” can both be true at the same time.
The painful part is that the people who lose one kind of job are not automatically the people who get the new ones. A displaced clerk does not instantly become an AI engineer. That gap — between disappearing tasks and newly valuable skills — is where unemployment, wage pressure and anxiety can grow.
AI often attacks tasks before it attacks occupations
Think about a receptionist. An AI system may answer routine questions, capture a caller’s details and schedule an appointment. But a real business still needs someone to handle unusual cases, angry customers, sensitive decisions, exceptions, accountability and the physical work around the service itself.
The same pattern appears in offices. Drafting, summarising, basic coding, document review, scheduling and routine customer support can be compressed dramatically. A company may need fewer people for some workflows. Or it may keep similar staffing and simply process far more work. Which outcome wins depends on costs, demand, competition and management decisions — not only on the raw capability of the model.
The first major pressure may be on entry-level work
This is one of the more uncomfortable possibilities. Entry-level workers traditionally learn by doing the repetitive work that senior workers no longer want to do: first drafts, basic analysis, data cleaning, simple support tickets, junior coding and document preparation.
If AI absorbs a large share of those beginner tasks, companies may ask a smaller number of experienced workers to supervise AI instead. That can improve productivity while quietly weakening the ladder that once created experienced workers in the first place.
That does not mean young workers have no future. It means the definition of “entry level” may move upward. Knowing how to use AI, verify it, combine it with domain knowledge and take responsibility for the result could become part of the minimum skill set rather than an advanced specialty.
Human value may shift toward judgment, trust and the physical world
AI is strongest where work can be represented digitally: text, images, software, spreadsheets, recorded speech and structured information. It is much less straightforward when the job involves unpredictable physical environments, human trust, legal responsibility, negotiation, leadership, caregiving, hands-on repair or complex coordination among people.
That is why plumbers, electricians, nurses, field technicians, salespeople, managers, teachers and many service workers are not simply “safe” or “unsafe.” Parts of their jobs can be automated while other parts become more valuable.
The IMF noted in 2026 that demand for new skills is rising fastest in professional, technical and managerial roles. In UK and US job postings, openings that mentioned a new skill tended to pay about 3% more, with larger premiums when several new skills were required. That does not prove everyone will earn more, but it does suggest that adaptation itself is becoming economically valuable.
Developing countries face a different AI problem
A 2026 ILO–World Bank study covering 135 countries found that developing economies generally have lower aggregate automation exposure than advanced economies, yet still have substantial potential for AI to augment work. The catch is infrastructure: slower digital adoption, weaker connectivity and fewer complementary investments can mean disruption arrives before the productivity benefits are widely shared.
In other words, a country can be “less exposed” to automation and still lose out if the most productive AI tools are concentrated elsewhere.
Could AI still cause serious unemployment?
Yes. There is no law of economics that says every transition must be painless. If AI capability improves faster than workers can retrain, if companies use it mainly to cut labour costs, or if new industries fail to absorb displaced workers quickly enough, unemployment can rise — especially in particular occupations, regions and age groups.
But the evidence available today does not justify treating permanent mass unemployment as an inevitable outcome. The more defensible conclusion is that AI creates a large transition risk: fewer openings in some roles, faster skill turnover, more pressure on wages and hiring, and potentially much higher inequality if the gains are concentrated.
Three futures can all begin with the same technology
Productivity dividend: AI helps people produce more, businesses grow, new services become affordable, wages rise and some workers gain shorter or more flexible hours.
Winner-take-most: a small number of firms and highly skilled workers capture most of the gains while routine knowledge work shrinks, bargaining power weakens and inequality rises.
Messy transition: some jobs vanish faster than new ones appear, unemployment rises for a period, education catches up slowly, and the labour market eventually reorganises around new tools.
Technology alone does not choose among those outcomes. Tax systems, education, competition, worker bargaining power, social protection, business incentives and how quickly new demand appears all matter.
What should a human worker do now?
The safest strategy is not to race AI at the exact tasks it does best. Learn to use it, then build value around the parts that still require context, judgment, accountability, taste, relationships or real-world execution.
That can mean pairing AI literacy with a domain: healthcare, manufacturing, finance, design, sales, law, maintenance, logistics, education or engineering. A person who knows a field deeply and can use AI well may be far more valuable than someone who only knows prompting — or someone who refuses the tools completely.
The future question is bigger than “Will AI take my job?”
A better question is: Which tasks will become cheap, which human capabilities become more valuable, and who owns the productivity gains?
If AI makes one worker capable of doing the work of three, society could get cheaper services, more output and new industries. It could also get fewer entry-level openings and a larger gap between people who control the tools and people whose tasks are automated. Both outcomes can exist at once.
So the honest answer is neither “AI will make everyone unemployed” nor “AI will create enough jobs automatically.” The future of work is being negotiated in real time — inside companies, schools, governments and millions of individual career decisions.
AI may change what a job is before it changes whether humans work at all.
Sources & further reading
- ILO (2025) — Generative AI and jobs: A 2025 update
- ILO (2026) — What AI exposure indicators tell us, and what they do not
- World Economic Forum — Future of Jobs Report 2025
- IMF (2026) — New Skills and AI Are Reshaping the Future of Work
- ILO–World Bank background study (2026) — Disruption without dividend?
