Analysis · 4 November 2026 · 6 min read
Should you change careers because of AI?
A high exposure score alone is no reason to quit. What retraining research shows, who has the most reason to move, and how to check where you are heading.
For most people, a high exposure score alone is not a reason to quit. It becomes a stronger case when most of your week is routine screen work, your savings are thin and your skills do not transfer easily. If you do move, the best evidence we have says the destination matters more than the course you take to get there.
How many people are actually doing this?
A sizeable minority say they are, though these numbers are soft. In a 2026 Campbellsville University survey of about 1,000 US adults, 14 percent said they were rethinking their career because of AI. In a 2025 survey of 1,000 Americans published by Southeastern Oklahoma State University, 11 percent said they had already pivoted in response to AI concerns and 22 percent were seriously considering it.
Those who moved reported transitions averaging about five months and retraining costs of about 1,300 dollars. Sixty eight percent said they felt more stable afterwards, and 27 percent reported impostor syndrome. All of this is self reported, and both universities sell programmes to people changing careers.
Does retraining work for people in exposed jobs?
The strongest evidence is a National Bureau of Economic Research paper by Hyman, Lahey, Ni and Pilossoph. It follows US government funded training from 2012 to 2024 and compares trainees with similar workers who only got help finding a job.
The result is encouraging. For workers leaving highly exposed occupations, the earnings gain from training rose from about 1,000 dollars a quarter before 2020 to about 3,000 dollars in 2022 to 2024. The authors attribute most of that to people moving into less exposed occupations, and a smaller part to training that builds skills which work alongside AI.
The destination effect shows up elsewhere too. Harvard's coverage of an earlier version reports a 29 percent earnings penalty for workers training toward highly exposed fields, compared with those taking general skills training.
The limits matter. These are mostly displaced workers on lower incomes, averaging about 40,000 dollars a year, in one public programme, and they lost their jobs for any reason, not only AI. One of the authors told Harvard that the results do not mean everyone should go and do a training programme. We found no equivalent study of mid career professionals leaving a job they still have.
Who has the most reason to move?
Research by Manning and Aguirre, published with Brookings, asks who could cope with losing a job, using savings, age, how far skills transfer and the size of the local job market. Exposure and capacity tend to go together. Of the 37.1 million US workers in the most exposed quarter of jobs, about 70 percent have above median capacity, including software developers, financial managers and lawyers. A later version of the paper uses slightly different totals but reaches about the same share.
The group to worry about is smaller: 6.1 million workers in the January version and 3.3 million in the later one. It is concentrated in clerical and administrative work, and in the January version about 86 percent are women.
Two of the largest of those jobs, as our index scores them: Office Clerks, General (58) and Receptionists and Information Clerks (60). If you are in one of these roles without much of a financial cushion, starting early costs less than starting after a layoff. The authors stress that exposure signals where change may come first and does not predict job loss.
Should you move now, or adapt where you are?
Three questions separate the two cases:
- Is the risky part most of your job? If AI reaches a quarter of your week, that part of the job may change while the rest stays. If it reaches most of your week, the job itself is in question.
- Is your employer already changing? Falling hiring for your role is a stronger signal than any forecast. Our post on layoffs shows which functions companies are cutting when they cite AI.
- Can you afford to wait, and to be wrong? A move takes months and has a cost, and a move made in a panic can land you somewhere no safer.
If you do move, where to?
Check where you are heading before you pay for anything. Moving from one exposed job into another gains little:
Someone leaving Customer Service Representatives (68) for Web Developers (61) has barely changed their position, and has paid for a course to do it. Compare Medical and Health Services Managers (38) or Registered Nurses (27), which sit much lower, although they come with their own licensing and training requirements.
A smaller move may be available too: shifting within your field toward the part that is harder to automate, such as from producing the work to managing it, or to dealing with the customer. It costs less, and you keep your experience. No study yet compares that directly with a full career change, so treat it as a judgment rather than a finding. Our post on which skills hold up covers what to move toward.
Our index scores 961 occupations, so you can compare your job with the ones you are considering. The assessment adjusts your score for how you actually work. It takes about two minutes, free, no account.
Where does your job sit?
Every number above is an occupation average. Your own exposure depends on your seniority, your sector and how much of your day is routine. Answer a few questions and get your personal score free.
Get my personal scoreFree · No credit card · No signup
Figures come from the AI Job Risk Index, the data behind our AI job risk calculator, 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.