Analysis · 25 August 2026 · 6 min read

AI layoffs in 2026: which jobs are being cut

AI was cited in US job cuts covering 205,000 workers by August 2026. How the cut categories compare with the research consensus, occupation by occupation.

2026 is the year layoff announcements started naming AI. Not hinting at it, naming it. By August, tracking by ResumePulse counted 205,000 US workers in cuts where AI or automation was cited as a driver, matching the whole of 2025 in under eight months, with May and June setting the highest monthly totals the tracker has recorded. Challenger, Gray and Christmas counted cuts approaching 49,000 positions in outsourcing adjacent functions by mid year alone.

A tracker tells you the size of the wave. It does not tell you whether the wave is landing where the research said it would. We maintain an index of occupation level AI exposure built from published research, so we can ask that directly: are the jobs being cut the jobs the evidence flagged?

Mostly, yes. And the places where the announcements and the research line up are worth more attention than the headline number.

Where the cuts are landing

Four categories keep recurring in the 2026 announcements: customer service, data operations, entry level software roles, and finance back offices.

Customer service is the clearest case. Reporting on the support industry describes Microsoft's support workforce falling from about 50,000 to roughly 40,000 in recent years, Uber cutting a tenth of its customer service operation, and Microsoft attributing around 750 million dollars a year in savings to AI in support. Forrester analysts now project that nearly half of current customer service jobs, 49 percent, will disappear by 2030.

Announcements are not measurements, and it pays to keep one grain of salt. A company that says AI when it means ordinary cost control is telling investors a better story, so some share of the 205,000 is rebranding. But the concentration is the signal: the same four functions, across unrelated companies, in the same months.

What the research said about these jobs

Here is the striking part. The four categories in the announcements sit almost exactly where the research consensus put its highest numbers. Our index currently scores Customer Service Representatives (68), Data Entry Keyers (63), and Bookkeeping, Accounting, and Auditing Clerks (58), each aggregated from multiple independent research sources.

Software is the more interesting split. The announcements talk about entry level engineering roles, and the research corpus distinguishes exactly along that line: Computer Programmers (59), the occupation defined by writing routine code to a spec, currently scores well above Software Developers (47), the occupation defined by designing systems and owning outcomes. The cuts are landing on the first job description, not the second.

They are also landing on the youngest people holding it. The Stanford payroll study Canaries in the Coal Mine found a 16 percent relative employment decline for workers aged 22 to 25 in the most AI exposed occupations, while their senior colleagues in the same occupations held steady. Our post on entry level jobs covers why the bottom rung takes it first: juniors do the concentrated version of the exposed tasks.

What the wave does not show

It does not show a broad labour market collapse, and honesty about that matters as much as the warnings. The cuts are deep and narrow: heavily concentrated in text and screen bound routine work, inside a handful of functions. Across the 20 research sources in our corpus, almost nothing clears the genuinely high risk threshold, a point the research agrees on more than the coverage suggests. Most occupations in the index sit in the middle band, where parts of the job automate and the job itself changes shape.

The 2026 wave is best read as the leading edge, the point where the most exposed task profiles stopped being a research finding and became a line item. If your work is mostly handling tickets, keying data, reconciling accounts or producing routine code to someone else's spec, the timeline stopped being theoretical this year. If it is not, the announcements tell you less about your job than the composition of your own week does.

Reading your own position

Start with your occupation's page in the index: the score, the band, and the specific studies behind it. Our statistics page keeps the headline numbers, ours and the published projections, current in one place. Then remember what an occupation average is. The people cut from support roles in 2026 were not average; they were the ones whose task mix matched what the tools had just learned to do. How to read your own job against the research walks through that logic, and the assessment below turns it into a number.

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.