The jobs AI is most likely to change first are the ones built on language and information: translating, writing, answering customers and selling. That is the finding of a Microsoft Research study that scored about 785 US occupations by how much of their work generative AI already helps with. But the 2026 layoff data tells a more cautious story: most job cuts this year were not about AI.
Want your own answer? Use our Will AI Replace My Job? check.
The 10 most AI-exposed jobs
Scores run from 0 to about 0.5 and measure how much of a job’s work generative AI can assist with, based on 200,000 real Copilot conversations.
| Occupation | AI applicability score |
|---|---|
| Interpreters and translators | 0.49 |
| Historians | 0.46 |
| Writers and authors | 0.45 |
| Sales representatives of services | 0.45 |
| CNC tool programmers | 0.42 |
| Broadcast announcers and radio DJs | 0.41 |
| Customer service representatives | 0.41 |
| Telemarketers | 0.40 |
| Political scientists | 0.39 |
| Mathematicians | 0.39 |
Source: Microsoft Research, Working with AI, occupation scores (CC BY 4.0).
And the jobs AI barely touches
At the bottom of the list, with scores close to zero, are hands-on jobs: orderlies, floor sanders and finishers, pile driver operators, rail-track equipment operators and foundry mold makers. Registered nurses and nursing assistants, electricians and maintenance workers also score low.
Exposed is not the same as replaced
The study measures where AI can help, not where jobs will be cut. Its authors point out that a high score can mean a job changes rather than disappears. The official projections show the same nuance: the Bureau of Labor Statistics expects employment of customer service representatives, one of the most exposed jobs, to fall 5% from 2024 to 2034, while still projecting about 341,700 openings a year (BLS Occupational Outlook Handbook).
What 2026 layoffs actually show
In our tech layoff tracker, 19 of 74 confirmed layoff announcements this year cited AI in the company’s own words, covering at least 6,240 of the 42,442 jobs with a disclosed number. They include Visa (about 2,600), Wix (about 1,000), Coinbase (700), HubSpot (about 660), Freshworks (500) and Angi (350). Outside tech, Barclays tied part of its 5,000 cuts to AI and Dow said its plan relies on AI and automation (see big company layoffs).
Recession-proof jobs? Look at low AI exposure and low layoff rates
No job is fully safe, but two signals help. Jobs with low AI scores are mostly hands-on, and some industries lay off far fewer people: over the last 12 months, government averaged 0.33% layoffs and discharges a month, finance and insurance 0.53% and health care 0.69%, against 1.08% for all US jobs (BLS JOLTS). Professional and business services (2.03%) and construction (1.84%) were the highest.
Check your own job
Our free AI and layoff risk check combines all three data sets for your job, your industry and your employer. Nothing you enter leaves your browser.
Data used in this article
- Tomlinson and others, Working with AI: Measuring the Applicability of Generative AI to Occupations (Microsoft Research)
- Bureau of Labor Statistics, Job Openings and Labor Turnover Survey
- BLS Occupational Outlook Handbook: Customer Service Representatives
- GoLayoffs Tech Layoff Tracker