Key points
- About 80% of US workers could have at least 10% of their tasks affected by large language models, and about 19% could see at least half their tasks affected (Eloundou et al., 2023). "Affected" means faster with AI, not gone.
- 1 in 4 jobs worldwide has some exposure to generative AI, but only 3.3% sit in the most exposed group, and clerical work is the most exposed (ILO, 2025).
- Employers expect about 170 million jobs to be created and 92 million displaced by 2030, and 39% of workers' core skills to change (WEF, 2025).
- On one AI assistant, 52% of conversations were people working with the AI and 45% were handing it a task to finish (Anthropic Economic Index, January 2026).
Paulo's new tool
Paulo is 44 and has handled insurance claims for 15 years. In March his team got an AI tool that reads a claim file and drafts the summary he used to write himself. It is good. Not perfect, but good.
The first week, he felt a little useless. The second week, he noticed something: the tool was fast at the summary, but it couldn't do the part where he calls a family after a fire, or decides that a strange claim is honest, or shows a new colleague what to look for. Then a friend sent him a headline about AI and jobs, and the question arrived: should I get out while I can?
Paulo isn't a real person, but his question is one many people are asking right now. Here is what the research says, and the check Paulo ran before deciding anything.
What the research says
of US workers could have at least 10% of their tasks affected by large language models. 19% could have half or more.
Eloundou et al., 2023jobs worldwide have some exposure to generative AI. Only 3.3% are in the most exposed group, and the ILO expects most jobs to change, not vanish.
ILO, 2025jobs employers expect to be created and displaced by 2030, with technology one of several drivers.
WEF, 2025These are estimates of what could change, made by people studying tools that keep moving. Treat them as a direction, not a forecast for your job.
The same WEF report lists creative thinking, resilience and flexibility, and curiosity and lifelong learning among the skills growing in importance, next to AI and data skills (WEF, 2025). The human parts of work are not shrinking. They are becoming more of the job.
Which jobs will AI change the most?
Clerical and office support jobs are the most exposed group in the ILO's global index (ILO, 2025). Employers in the WEF survey expect the biggest declines in clerical roles such as bookkeeping and data entry clerks, and growth in nursing, teaching and care work (WEF, 2025). Some highly digital professional and technical jobs also score higher, but in most jobs exposure means AI doing part of the work alongside the person, not instead of them.
Ask "which tasks?", not "which job?"
Most jobs are a bundle of tasks, and AI touches the bundle unevenly. That's why the research above measures tasks, not job titles.
| Tends to change a lot | Tends to change some | Tends to change less |
|---|---|---|
| First drafts of text and code | Analysis and reports | Hands-on and physical work |
| Summaries, translation, data entry | Customer service and sales | Care, teaching and coaching in person |
| Scheduling and routine admin | Design and research | Leading people through change |
| Searching for information | Planning and project work | Judgment calls where someone is accountable |
Most jobs have tasks in all three columns. Here is what that looks like in four common jobs:
| Job | Tasks AI can draft | Tasks that still need a person |
|---|---|---|
| Insurance claims analyst | Claim summaries, standard letters | Calls with families, judging unusual claims, training new staff |
| Teacher | Quiz questions, lesson outlines, report-card wording | Reading the room, helping one student who is stuck, talking with parents |
| Software developer | Boilerplate code, tests, documentation drafts | Deciding what to build, reviewing what AI wrote, owning what ships |
| Marketing coordinator | Ad copy variants, social posts, meeting notes | Choosing the message, working with clients, approving the final version |
These are examples to show the pattern, not a ranking of jobs. The question is the mix in your week.
A 30-minute check of your own job
Do this before you read another headline. It turns a vague worry into a list you can act on. Here's how Paulo ran it.
1. List last week's tasks
Write down 15 to 20 things you actually did. Use your calendar and sent emails to jog your memory.
Paulo: 18 tasks, from "draft claim summaries" to "train the new hire".
2. Mark each one
Three labels: AI can draft it, AI can help with it, or it needs you (a relationship, a judgment, your hands).
Paulo: 7 draft, 6 help, 5 need him.
3. Look at the first column
That time is likely to free up. Decide what you'd like to fill it with, before someone else decides for you.
Paulo: about a day a week, freed from writing summaries.
4. Look at the third column
Those tasks are worth getting better at. They often point to a direction if you do decide to move.
Paulo: the hard calls with families, odd claims, and teaching.
5. Pick one move for the next three months
Grow a third-column task inside your job, or test a direction outside it.
Paulo: he asked to take on complex claims and to train new staff.
If most of your week sits in the first column, that's a good reason to plan a gradual shift while you're employed, not to quit in a hurry. Our guide on how to change careers without quitting shows the routes that keep your salary.
If you do decide to switch
Pick a field where AI is mostly adding tools, not removing the entry-level work. We cover this in depth in our guide on changing careers at 40: a three-question test for any career, examples of where junior work is shrinking, and the retraining trap to avoid. The short version:
- Is most of the day predictable screen or text work? Expect AI to take a growing share.
- Does it need presence, trust, real-stakes judgment or hands? Those parts are harder to hand to a machine.
- Are people in the field doing more with AI, or being replaced? Ask five of them.
Don't decide on headlines
Headlines are about jobs; the research is about tasks. A job where half the tasks change can still be a better job afterward, if you move toward the half that needs you.
How each type tends to adapt
These are tendencies, not rules. Most people are a main type plus an accent.
Peak: Tends to adopt new tools early when they help hit a goal. Keep the decision itself: let AI speed up the work, not make the call you're accountable for.
Horizon: Tends to enjoy trying new tools and often finds new uses for them. The challenge is picking one use and finishing it.
Campfire: Tends to spend much of the week on people work, which changes less. Let AI take some admin off the plate and keep the time for conversations.
River: Tends to bring careful judgment. Checking AI output and spotting what's wrong or missing are growing parts of many jobs.
Root: Tends to turn a new tool into a reliable process. Expect to revise that process more often than usual, since the tools keep changing.
If Paulo took the test
Riverwith a Root accentA River with a Root accent fits Paulo's choice. He didn't jump. He found the part of his work that needs careful judgment, and made it the center of his week. A Horizon in the same spot might have become the team's go-to person for the new tool. Both are good moves; they just start from different strengths.
Paulo's scores are made up to show what a result looks like.
Lu's tip: Run the 30-minute check again in six months. The tools change fast, and so will your three columns.
Questions people ask
Will AI take my job?
For most people, AI is more likely to change parts of the job than remove it. The most exposed tasks are routine writing, data and admin. The 30-minute check above shows how much of your week that is.
Should I change careers because of AI?
Not on headlines alone. If most of your tasks are in the "AI can draft it" column, plan a gradual shift toward work with more judgment, people or hands-on skill, and test it while you're employed.
Which jobs are safest from AI?
No list is certain. Work that combines physical skill, in-person care or accountable judgment tends to change less, but most of those jobs will still use AI tools somewhere. The three-question test in our career change guide helps you judge any field.
Should I learn AI instead of changing careers?
It's not either-or. Learning to use AI well in your current job can be the cheapest career change of all, and it makes any later switch easier. Use it for real work every week, not just to try it.
Are entry-level jobs more at risk than senior ones?
Junior roles are often heavy on the first column: drafts, summaries, data entry. That's why retraining into a junior role in a field where AI is absorbing that work is a trap. Experience, judgment and relationships are harder to replace.
Can a personality test tell me which tasks to grow into?
No test can choose for you. A good one describes what tends to give you energy and what drains it, which helps you pick which third-column tasks to lean into. Use it next to your skills and what you learn from trying.
Sources
- Eloundou, T., Manning, S., Mishkin, P., and Rock, D. (2023). GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv:2303.10130. Later published in Science, 2024.
- Gmyrek, P. et al. (2025). Generative AI and Jobs: A Refined Global Index of Occupational Exposure. ILO Working Paper 140.
- World Economic Forum (2025). The Future of Jobs Report 2025.
- Anthropic (2026). Anthropic Economic Index report, January 2026.
You walk a trail instead of filling a form
Tipolu is a short 3D walk. At each crossroads you pick the path you'd really take, and the companion walking with you at the end shows who you are.
12 to 20 crossroadsEveryday choices on a trail, like a market, a lake, a broken well.
Two campfire questionsShort open answers in your own words.
Your mix, not a boxA main type, an accent, and 0 to 100 scores on five traits, with a report that includes work.
Try three crossroads from the real walk
Pick the path you would really take. No right answers.
In an empty village you find an old workshop full of odd tools and half-built machines.
A storm keeps your group in a valley potter's workshop all day, and she lets you use her wheel.
A baker hires your group for the morning rush. The line is long, and many customers are regulars.
Three forks are a guess. The full walk uses 12 to 20 of them plus two campfire questions, and shows your main type, your accent and what that means for work.
Play the free test · 5 min
Which one is walking with you?
Find out how you tend to work before you pick your next path. A short walk, 12 to 20 crossroads, and a report that includes work.
Play the free testAbout 5 minutes · no signup · free