What Jobs Will AI Replace by 2030? What the Forecasts Actually Say

The most-cited answer to this question comes from the World Economic Forum, and it is not the one the headlines usually run with. Its Future of Jobs Report 2025 surveyed employers representing millions of workers and projected that by 2030, around 170 million new jobs will be created and around 92 million displaced — a net gain of roughly 78 million.

That is a forecast of churn, not collapse. It also averages across the whole world economy, which means it says almost nothing about your specific job. The useful detail is in which categories move, and in which direction.

The roles employers expect to shrink

The WEF report is consistent about the shape of the decline: it is concentrated in clerical and routine information handling, not in any one industry.

• —Data entry clerks — the most consistently cited declining role across every recent forecast.
• —Administrative assistants and executive secretaries — expected to fall the most in absolute numbers, alongside cashiers.
• —Cashiers and ticket clerks.
• —Postal service clerks and bank tellers.
• —Accounting, bookkeeping and payroll clerks.
• —Graphic designers, which is newer to the list and is specifically attributed to generative AI.

Two things stand out. The first is that this is largely a continuation of a trend that predates generative AI — bank tellers and postal clerks have been declining for two decades for reasons that have nothing to do with language models. The second is that graphic designers appearing on the list is genuinely new, and is the clearest signal that the current wave reaches work previously considered safe because it was creative.

The roles expected to grow

The same report projects growth in roles that are physical, care-based, or technical: farmworkers, delivery drivers, software developers, care and health workers, and the fastest-growing category in percentage terms, AI and machine learning specialists.

That growth column matters more than it usually gets credit for. "AI will replace jobs" and "AI will create jobs" are both true, and the gap between them is not evenly distributed — it lands on different people, in different places, at different times. A net gain of 78 million is no comfort at all if you are one of the 92 million.

How much should you trust a 2030 forecast?

Less than the confident framing implies, and here is why in plain terms.

1. 01It is a survey of employer expectations, not a measurement. It records what executives think will happen, and executives have been wrong about technology timelines in both directions many times.
2. 02Five-year forecasts in fast-moving fields are weak. The same exercise run in 2019 did not anticipate what 2023 looked like.
3. 03"Displaced" is not "unemployed". A displaced role often becomes a changed role at the same employer, which is invisible in headline numbers.
4. 04The category is not the person. "Accounting clerks decline" and "this particular accountant does better than ever" are entirely compatible, and the research below shows exactly that. The field study in will AI replace accountants is the clearest example.

The most rigorous work on task exposure is a Microsoft Research paper, Working with AI, which mapped 200,000 anonymised AI conversations onto occupational task databases. It found high applicability across information work. But the authors then published a follow-up note specifically to say the study does not support displacement conclusions, and that "a job is far more than the collection of tasks that make it up". Anyone citing that paper as a list of doomed jobs is misusing it. We went through that distinction in detail in What Jobs Are Safe From AI?.

What has actually happened so far

Forecasts are cheap. Here is the measured record, which is smaller and more specific than the projections.

Challenger, Gray & Christmas, which tracks announced US job cuts and the reasons employers give, attributed 54,836 cuts to AI in 2025 — bringing the total since 2023 to 71,683, which means last year alone exceeded the two before it combined. Real, rising, and still a small share of overall layoffs. It also counts stated reasons, and "AI" is a more comfortable public explanation than "we over-hired" — AI and job loss goes through what that count does and does not measure.

On the other side, PwC's 2026 Global AI Jobs Barometer analysed close to a billion job postings across 24 countries and found roles requiring AI skills carry a 62% wage premium over comparable roles without them, up from 57% a year earlier, with postings for those roles growing 69% against 9% for the wider market.

The measured effect so far is less about jobs disappearing and more about the same job paying differently depending on whether you can use the tools.

A more useful question than "will my job exist"

Nobody can tell you whether your title survives to 2030. What you can work out today, in about ten minutes, is your task exposure:

1. 01List what you actually did last week — tasks, not your job title.
2. 02Mark each one that is producing, summarising, reformatting, or retrieving information. Those are the exposed tasks.
3. 03For the top two or three, learn to do them with AI to a standard where the output needs editing rather than rewriting.
4. 04Keep the parts that involve judgement, accountability, and relationships. Those were always the job.

Step three is the one people skip, usually because a bad first attempt convinced them the tools do not work. Most of the time the request was the problem — How to Write a Prompt That Works on the First Try covers that, and What AI Is Actually Bad At covers the failure modes worth knowing before you rely on any of it.

The short version

The best available forecast expects more jobs created than lost by 2030, with the losses concentrated in clerical and routine information work. Treat the number as a direction of travel, not a prediction about you. The measured evidence to date shows a modest displacement effect and a large and growing pay gap in favour of people who can use the tools — which is the part you can do something about this month.

Coursium teaches that practical use, and the checking that makes it safe to rely on. Stay ahead of AI by learning the tools on your phone.