No: current evidence does not show that AI is about to eliminate everyone’s job. But the history of automation is not a promise that every worker will be protected, either. Past changes show how job losses, new work and productivity gains can happen at the same time. Today’s estimates measure which work could be affected—not how many people will be laid off.
What does “AI exposure” actually mean?
An occupation is considered exposed when some of its tasks could be affected by generative AI. That is different from predicting that the occupation—or the people doing it—will disappear. A tool may take over some tasks, change how others are done, or help workers do them faster while leaving the job itself in place.
The International Labour Organization’s 2025 update estimates that one in four workers worldwide is in an occupation with some degree of generative AI exposure. Its assessment covers nearly 30,000 tasks and concludes that most exposed jobs are more likely to be transformed than made redundant. The “one in four” figure is an estimate of exposure, not a forecast that one in four workers will lose a job. ILO, “Generative AI and jobs: A 2025 update”.
Exposure estimates also depend on the method and period used. The ILO’s 2025 mean automation score of 0.29, compared with 0.30 in its 2023 estimate, is a score from a revised exposure methodology—not a measured job-loss rate. The standard deviation likewise changed, from 0.30 to 0.14. These figures describe the distribution of scores under the respective methodologies, not the share of jobs that will be eliminated. ILO, 2025 update.
#1 Best Overall
What happened in earlier automation waves?
Technology has repeatedly changed the tasks people do and the kinds of work employers need. That history matters because it shows why counting tasks that machines might perform is not enough to predict what happens to employment overall: automation can reduce demand for some work while new tasks, occupations and productivity gains push in other directions.
An OECD analysis published in 2024 examined 21 countries and 38 occupations over 2012–2019, focusing on places and occupations previously classified as at high risk of automation. It found no support for net job destruction at the broad country level in that analysis. The finding does not show that nobody lost work, that every occupation fared well, or that future AI effects will follow the same pattern. National employment totals can obscure losses concentrated in particular jobs, communities or groups of workers. OECD, “Technological change and the labour market: Megatrends and the Future of Social Protection”.
Rank #2
That is why “we’ve been here before” is useful as a comparison, not as reassurance that the outcome is settled. A 2017 ILO brief on automation describes job creation and destruction as processes that can occur together and highlights inequality and the difficulty workers may face moving from old jobs to new ones. ILO, “New automation technologies and job creation and destruction dynamics”.
How can AI change jobs without simply replacing them?
The OECD identifies three channels through which AI affects labour markets: “AI affects labour markets through three main channels: i) automation of existing tasks, ii) creation of new tasks and occupations and iii) improving productivity.” OECD, “Skills in the AI age,” executive summary, 2026.
- Automation: AI can perform tasks that workers previously did, potentially reducing the need for some roles or changing the skills those roles require.
- New work: Employers may develop new tasks and occupations as technology is adopted.
- Productivity: AI may let workers or firms produce more with the same resources. The effect on employment depends on what happens next, including whether higher output or new activities create demand for workers.
These channels can overlap. A role may lose some tasks and gain others; a company may need fewer people for one activity while hiring for another. The net effect depends on how these changes balance, so exposure alone cannot settle whether a particular job will grow or shrink.
Which workers may face greater risk?
Exposure to AI and the likelihood of automation are not the same thing. The OECD notes that some high-skill work can be highly exposed to AI while remaining less likely to be automated because it involves non-routine cognitive and social skills. By contrast, routine manual or cognitive work in low- and middle-skill jobs can face higher automation risk. These are broad patterns, not a prediction for an individual worker or occupation. OECD, “Skills in the AI age,” 2026.
Adoption is rising, but adoption statistics are not job-loss statistics. OECD data indicate that the share of firms using AI in OECD countries rose from around 7% in 2021 to 20% in 2025. The OECD also estimates that around one-quarter of workers were exposed to generative AI in 2022–2024. That worker estimate is similar in broad scale to the ILO’s 2025 figure, but the methods and reference periods are not identical. Neither figure says how many jobs have been or will be eliminated. OECD, 2026.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why job totals do not tell the whole story
Even if employment does not fall overall, the transition can be difficult for workers whose tasks or roles change. Some may need to move to different work; the new work may be in another place, require different skills or offer different conditions. Gains and losses can also be distributed unevenly, so a stable national job total does not mean everyone benefits.
Best Value
Job quality matters, too. The ILO’s analysis considers how AI can affect working conditions through algorithmic management, as well as the labor involved in producing AI systems. That makes the question broader than whether the technology creates or destroys a net number of jobs: it also concerns who gains, who bears the costs and how work is managed. ILO, “Artificial intelligence adoption and its impact on jobs,” 31 May 2025.
What should workers take from the evidence?
Past automation offers a reason not to treat a list of exposed tasks as a list of doomed careers. It does not justify assuming that every transition will be smooth or that future employment must rise. The strongest conclusion is narrower: technology changes tasks and can affect workers unevenly, while aggregate job outcomes depend on several forces operating at once.
For someone assessing their own work, the useful question is not simply “Can AI do part of my job?” It is how much of the role consists of automatable tasks, what judgment or interaction remains, and whether the occupation is likely to change as employers adopt the tools. The available economy-wide estimates do not answer that for any one worker.
In 2023, the OECD said the then-available aggregate data showed no signs of slowing labour demand yet, while also noting the difficulty of detecting AI’s effects in the data. That observation was bounded by the data available at the time; it is not a guarantee about future labour demand. OECD, “Artificial intelligence and jobs: No signs of slowing labour demand (yet),” 2023.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
