Will AI take my job? How to think about your own

Think in tasks, not job titles

AI does not take a job title. It takes over, or speeds up, particular tasks. A job is a bundle of tasks, so the useful question is how much of your bundle is the kind of work AI is already doing.

Write down what you actually did last week, in rough hours. Then sort each task:

If most of your hours sit in the first row, your job is more exposed. That does not mean it disappears. It usually means the job changes: less time on first drafts, more on checking, deciding and dealing with people.

What the evidence says so far

These are the figures we checked at their source on October 2, 2026.

• Usage. Anthropic measures which tasks people actually do with its AI model, Claude, at work. It calls this "observed exposure". The most exposed occupation was computer programmers, at 75%. At the other end, "30% of workers have zero coverage", because their tasks barely appear in the data (March 2026).
• Unemployment. In the same study: "We find no systematic increase in unemployment for highly exposed workers since late 2022."
• Young workers. Stanford's Digital Economy Lab found employment of 22 to 25 year olds in highly AI-exposed jobs about 19% below where it would be had it kept pace with their peers in less exposed jobs, as of June 2026 (15% a year earlier). Experienced workers show no comparable gap. The authors call these descriptive patterns, not proof of cause.
• Layoffs. Employers cited AI in 120,136 announced US job cuts from January to September 2026, about 21% of the total and the leading stated reason this year. All announced cuts were down 39% on the same months of 2025 (Challenger, Gray and Christmas).
• Worry. 27% of US workers worry technology could soon make their job obsolete, a record (Gallup, August 2026; 507 workers, margin of 6 points).
• Projections. The Bureau of Labor Statistics projects US employment to grow 3.5% from 2025 to 2035. Its biggest projected decline is office and administrative support, partly from "automation tools, including those powered by AI".

How exposed is your occupation?

Anthropic's score per occupation, next to the official ten-year projection. Each page has the detail: the tasks, what people use AI for, and what to lead with on a resume.

A 0% means the job's tasks appeared too rarely in Anthropic's data to count, not that AI cannot touch them.

Two things stand out. Exposure and job growth do not move one for one: software developers are among the more exposed and still projected to grow. And Anthropic found only a mild link overall: for every 10 points more exposure, the BLS growth projection was 0.6 points lower.

If you are weighing a change of field, jobs less exposed to AI lists hands-on and care work with the same two figures.

What the numbers cannot tell you

• Usage data measures how people use AI tools, not job losses. A task showing up a lot in Claude conversations means people use AI for it. It does not mean anyone was let go.
• Projections are estimates. BLS itself warns that the precision of its figures "does not account for the inherent uncertainty of predicting long-term changes in the labor market."
• Layoff counts are what companies say. An announcement that cites AI is not proof that AI did the work.
• Nobody knows the answer for certain. Not us, not the researchers, not the people selling scores.

What to do with this

1. Lead with the less exposed part of your work. On your resume, put first the decisions you made, the people you worked with or led, and the results you were responsible for. Move pure production (drafted, entered, compiled) lower or fold it into a result.
2. Show that you use the tools. One honest line about how you used AI to do a task faster, and what you checked, reads better than either silence or a list of buzzwords.
3. Check how your resume reads. The resume check on this page is free and needs no account. It reads your actual resume and shows what an employer's software and a recruiter will see. It does not predict your job's future. It shows whether your experience comes across the way you mean it.