What the Job Forecasts Actually Say About AI

Is AI going to take your job? The two institutions most often quoted on the question both published in 2026, and neither gives a yes or a no. The World Economic Forum offers four scenarios instead of a forecast. The World Bank offers percentages that change sharply depending on which country you work in. What follows is what the documents say, what they do not say, and how to read the next headline built on them.

What did the World Economic Forum publish?

In January 2026 the Forum released Four Futures for Jobs in the New Economy: AI and Talent in 2030. It is a scenario exercise, not a prediction. It crosses two uncertainties, how fast AI advances and how ready the workforce is, and describes the four worlds that result.

Scenario	AI progress	Workforce	What happens to work
Supercharged Progress	Exponential	Ready	Many jobs disappear, new occupations scale quickly, safety nets and governance lag behind
The Age of Displacement	Exponential	Not ready	Automation outruns reskilling, unemployment spikes, social trust erodes
Co-Pilot Economy	Incremental	Ready	Augmentation rather than mass automation, after an AI bubble bursts
Stalled Progress	Incremental	Not ready	Uneven gains, routine roles hit first, skilled trades rise in value

The most revealing numbers in the report are not about the future at all. They come from the Forum's survey of more than 10,000 executives, who were asked what they expect AI to do. About 54% said it will displace a large number of existing jobs. About 24% said it will create a large number of new ones. Nearly 45% expect higher profit margins. Only 12% expect higher wages.

The people deciding how AI is deployed expect margins to rise far more often than they expect wages to.

That is not a forecast of what the technology can do. It is a statement of intent from the people who will choose what to do with it.

What about the 170 million new jobs?

This is the figure most often used to end the conversation: 170 million jobs created by 2030, 92 million displaced, a net gain of 78 million. It comes from the Forum's Future of Jobs work, which the Forum itself still cites.

Three things are worth holding on to before repeating it.

What did the World Bank find?

The World Bank's report, published on 4 August 2026 and summarised by Reuters, asks a different question: where in the world are jobs exposed?

	High-income countries	Low- and middle-income countries
Jobs threatened by generative AI	14.2%	4.5%
Jobs expected to see meaningful productivity gains	18.7%	16.2%

On these numbers generative AI is roughly three times more likely to threaten a job in a rich country than in a poor one, while the potential upside is about the same in both. The Bank's conclusion is that developing economies have more to gain and less to fear, provided they close gaps in electricity, connectivity, devices and digital skills. The same report warns that AI could also bring greater income inequality, harder-to-detect misinformation and political repression.

Is the picture really that reassuring for poorer countries?

Not entirely, and the caution comes from the Bank's own background research. A joint paper by the International Labour Organization and the World Bank, published in March 2026 and covering 135 countries and about two-thirds of global employment, found a timing problem.

In lower-income countries, the workers whose jobs are most exposed to automation are often already online. The workers who could gain the most from AI often are not. So the losses can arrive quickly while the gains wait for infrastructure. The exposed jobs are also disproportionately the good ones: clerical and administrative roles that have been a route into stable, decent work, particularly for women and young people.

A low exposure percentage, in other words, can describe a country where few jobs are at risk and the ones that are happen to be the ladder.

What do these reports not tell us?

• Exposure is not job loss. A job being exposed to generative AI means the technology could do some of its tasks. Whether anyone is dismissed depends on employers, law, bargaining power and demand.
• Scenarios are not probabilities. The Forum does not say which of its four futures is most likely.
• Surveys measure expectations. Executives expecting displacement is evidence about plans and mood, not about outcomes.
• None of these documents can tell an individual what will happen to their job. They describe distributions, and nobody lives in a distribution.

How should you read the next jobs headline?

Read together, the reports support neither panic nor comfort. They say the effect of AI on work is still largely undecided, and that the deciding is being done now, mostly by people who expect to come out ahead.