Computer programmers and software developers often share an office. They write code, use the same tools and answer to the same boss.
Yet on paper, their futures split in two. Programmers are among the jobs most exposed to AI, and the US government expects 11% fewer of them by 2032. Developers are exposed too, but their numbers are forecast to grow by 26%.
That gap is the real story behind every "jobs AI will replace" list. AI exposure alone doesn't decide who loses work. What matters is exposure plus demand. So we took the task-by-task ratings for 923 US jobs, set them against official job forecasts, and sorted them into three groups: the jobs at real risk, the jobs AI is reshaping, and the jobs it can barely touch.
The short version
| Jobs at risk | Jobs being reshaped | Jobs AI can't replace (yet) | |
|---|---|---|---|
| AI exposure | High | High | Low |
| Forecast to 2032 | Shrinking | Growing | Growing |
| Examples | Executive assistants, customer service, order clerks | Data scientists, security analysts, software developers | Physician assistants, home health aides, cooks |
| What to do | Move toward tasks AI is weak at | Learn the tools before others do | Use AI for the paperwork around your work |
Want your own number? The Will AI take my job? checker shows your job's score, where it ranks and which of your tasks still need a human.
Reader poll: How worried are you about AI taking your job? Vote on GeekSpark
Which jobs will AI replace first?
These are the jobs where two warning lights flash at once: AI can speed up more than half of the work, and the Bureau of Labor Statistics (BLS) already expects fewer of them. All have at least 100,000 US workers.
| Job | AI exposure | Forecast, 2022 to 2032 |
|---|---|---|
| Legal secretaries | 76% | −22% |
| Executive assistants | 74% | −21% |
| Customer service representatives | 70% | −6% |
| Advertising sales agents | 69% | −7% |
| Computer programmers | 68% | −11% |
| Editors | 65% | −4% |
| Order clerks | 60% | −18% |
| Secretaries and admin assistants | 57% | −12% |
Notice the pattern. These jobs move information around: typing it, routing it, summarizing it, answering questions about it. That's the work language models handle best.
Customer service stands out for its size. Nearly 3 million Americans do it, and it scores 70% for AI exposure. Even a small shift there touches a lot of people.
The jobs AI is changing, not killing
High exposure doesn't always mean fewer jobs. In these roles AI can do much of the work, yet demand is growing fast:
| Job | AI exposure | Forecast, 2022 to 2032 |
|---|---|---|
| Data scientists | 59% | +35% |
| Information security analysts | 54% | +32% |
| Software developers | 45% | +26% |
| Operations research analysts | 63% | +23% |
| Software testers (QA) | 61% | +20% |
| Logistics analysts | 65% | +18% |
What do they share? The world wants a lot more of what they produce: more software, more security, more analysis. When AI makes a job faster and demand keeps rising, companies tend to ask for more work, not fewer people.
That's also why programmers and developers part ways. Programmers mostly turn finished designs into code, a narrow task AI now does quickly. Developers decide what to build, how the pieces fit and what can go wrong. AI speeds up both jobs, but only one of them grows with demand.
Jobs AI can't replace (for now)
At the other end are jobs where AI barely registers and demand is rising anyway:
| Job | AI exposure | Forecast, 2022 to 2032 |
|---|---|---|
| Physician assistants | 12% | +27% |
| Physical therapist assistants | 10% | +26% |
| Home health and personal care aides | 4% | +22% |
| Restaurant cooks | 12% | +20% |
| Medical assistants | 14% | +14% |
| Highway maintenance workers | 0% | +8% |
They all happen in the physical world, with real people, in places that change minute to minute. A chatbot can write the care plan. It can't help someone out of bed.
That doesn't make these jobs AI-free. The paperwork around them, such as notes, scheduling, quotes and billing, is exactly what AI speeds up. The people who hand it that admin get more time for the part only they can do.
Is AI actually causing layoffs right now?
Yes, but less than the headlines suggest, and the trend just shifted.
Employers named AI as the reason for 121,401 US job cuts in 2026 through October, according to Challenger, Gray & Christmas data reported by HRD America. That's about one in five of all announced cuts this year, and more than double the 54,836 blamed on AI in all of 2025.
October tells a different story. Only 1,265 cuts that month were put down to AI, the lowest monthly figure of the year. And not everyone takes the AI label at face value. Responding to claims that AI is replacing workers en masse, Lightcast economist Elena Magrini said plainly: "we are not seeing that in the labor data."
The honest reading: some companies are cutting roles because of AI, and others are using "AI" as a tidier word for cost cutting.
What "exposed to AI" really means
The scores come from a 2023 study by researchers at OpenAI and the University of Pennsylvania. They took the official task list for every US occupation (from O*NET, the Labor Department's job database) and asked one question about each task: could an AI language model do it in half the time, at the same quality?
- If a chatbot alone could do it, the task counts fully.
- If it needs extra software built on AI, it counts half.
- If not, it doesn't count.
A job's score is the share of its tasks that qualify. Across all 923 jobs the middle score is 31%. Only about one in five jobs scores 50% or more.
Two limits are worth knowing. The ratings are from 2023, and AI has improved since, so real exposure may be a bit higher today. And "exposed" means AI can speed up the task, not that it takes the whole job. Most jobs are a bundle of tasks, and AI usually takes some of the bundle, not all of it.
How to protect your job from AI
- Find out where you actually stand. Look up your job in the AI job checker. Guessing tends to swing between panic and denial.
- Pick one exposed task and hand it to AI this week. Learn where it saves you an hour and where it confidently gets things wrong. Spotting those mistakes is a skill on its own.
- Grow the part of your job AI is weak at. Judgment, trust, persuasion, hands-on work and messy real-world problems are the hardest things to automate.
- Write down what you do with the time you save. "I closed 30% more tickets" is a much stronger line at review time than "I used ChatGPT."
- If your job is in the at-risk group, start moving now. Look for a growing job next door that uses the same skills. Executive assistants score 74% and face a 21% drop; project management specialists, who also plan, schedule and keep people on track, score 40% and are forecast to grow 6%.
For the wider debate, including the power and water AI runs on, read why AI should be a tool, not a human replacement.
Reader poll: Has AI already changed how you do your job? Vote on GeekSpark
Quick answers
What jobs will AI replace first?
Office jobs that move information around and are already shrinking: legal and executive assistants, customer service representatives, order clerks and computer programmers. They combine high AI exposure (57% to 76%) with falling job forecasts.
What jobs can't AI replace?
Hands-on, in-person work: physician assistants, home health aides, cooks, and skilled trades such as electricians and HVAC technicians. They all score 15% or less for AI exposure, and most are forecast to grow faster than average.
Will AI replace software engineers?
Not on current evidence. Software developers score 45% for AI exposure, but their jobs are forecast to grow 26% by 2032. Computer programmers, who mostly write code to a spec, are forecast to shrink 11%.
How many jobs has AI replaced in 2026?
US employers blamed AI for 121,401 announced job cuts through October 2026, about 20% of all cuts, based on Challenger, Gray & Christmas data. Only 1,265 of those came in October.
How can I tell if my job is at risk from AI?
Check two things: how much of your job AI can speed up, and whether demand for your job is growing. The free AI job checker shows both for 923 US jobs.
Sources
- Eloundou, Manning, Mishkin and Rock: GPTs are GPTs (2023)
- GPTs are GPTs: occupation and task data
- US Bureau of Labor Statistics: Employment Projections
- HRD America: AI blamed for 121,401 job cuts (Challenger, Gray & Christmas data)
- O*NET OnLine: occupation tasks
Originally published on GeekSpark.
Top comments (0)