Guide · 25 August 2026 · 6 min read

When will AI take your job?

There is no single date. What the 2026 layoff wave and published research actually say about timing, and why your task mix matters more than your job title.

Anyone who answers this question with a single year is selling certainty that does not exist. The research does not produce a date. It produces something more useful once you stop expecting a date: a schedule, ordered by the kind of work, that is already partly underway.

So the honest answer has three parts. For a narrow set of task profiles it is already happening. For most occupations the published projections cluster around gradual transformation through 2030 and beyond, not disappearance. And which of those describes you depends on your task mix far more than on your job title.

Part one: for some work, the answer is now

This stopped being a forecast in 2026. By August, AI was cited in US job cuts covering 205,000 workers, concentrated in customer service, data operations, entry level software and finance back offices. Our analysis of the 2026 layoff wave compares those categories with the research consensus occupation by occupation; the short version is that the cuts are landing where the evidence said the pressure was highest.

The other measured result is about who goes first inside an exposed occupation. Stanford's payroll data study Canaries in the Coal Mine found workers aged 22 to 25 in the most exposed occupations already down 16 percent in relative terms, while senior people in the same occupations held steady. When arrives earlier the more junior you are, because juniors hold the concentrated version of the automatable tasks.

Part two: what the published timelines actually say

The serious projections do carry dates. Read together, with their methods in view, they describe a decade of reshaping rather than a cliff.

By 2030. Forrester projects that 49 percent of current customer service jobs will disappear by 2030. McKinsey's scenarios put roughly a third of current work hours in reach of automation by 2030 in their faster cases. The World Economic Forum's Future of Jobs survey expects employers to shed 92 million roles and create 170 million new ones by 2030, a net gain, though not for the same people doing the same tasks.

The famous big number. Goldman Sachs estimated in 2023 that generative AI could expose work equivalent to around 300 million full time jobs globally. Exposure there means tasks touched, not jobs deleted, which is how a startling number and a gradual labour market can both be true.

Why they disagree. These studies are not measuring one thing. Some measure what language models could do with a job's tasks, others what machines could absorb in principle. The two families barely correlate, so a timeline built on one will always look wrong through the lens of the other. None of this is a defect in your favourite study; it is a reason to never trust a date that arrives without a method.

Part three: why capability does not set the date

The models improving is not the clock. Adoption is, and adoption runs on economics. An MIT study of computer vision automation found that at current costs, only about a quarter of the wages attached to technically automatable vision tasks would be economical to automate today. Capability arrives first; rebuilding workflows, integrating systems, handling liability and getting the cost below the wage all come later, and they come at different speeds in different industries.

That is why the question has no single date even in principle. The same task automates years apart at a bank and at a startup. Watch adoption in your own function, not model announcements: when the standard tools of your field ship features that do your tasks rather than assist them, the clock you care about has started.

Turning the question on your own job

The unit that has a timeline is not the occupation, it is the task. Routine text and document work goes first, and 2026 shows that phase running. Judgment calls, relationships and physical work sit later on the schedule or on no visible schedule at all. Your occupation's baseline across the 961 roles in our index tells you the floor; how much of your own week is routine tells you where you sit against it.

The personalised report goes one step further and puts an estimated displacement window, in years, on your specific role and task mix. It is the closest thing to an honest answer to when that the evidence supports: a range, for your situation, that moves as the research does.

Figures come from the AI Job Risk Index and were current when this was published. Scores change as new research is ingested, so the index is always the live version. See how scoring works. Informational guidance based on published research, not professional career or financial advice.