You did everything right, and the first rung of your career is starting to disappear. Which jobs AI is taking now, and what that means for you.
This in-depth guide covers everything you need to know about jobs ai will replace: who is losing work first. Based on verified income data and real-world case studies from our database of 138 side hustle tactics.
AI is already replacing parts of many jobs and the whole of a few: customer service, data entry, telemarketing, translation, writing, basic coding and office administration are where the cuts show up first. Most jobs are being hollowed out task by task, and the clearest damage so far falls on people trying to get their first job in an exposed field. In 2026 AI became the most cited reason for announced US job cuts, and the occupations most exposed now reach into work people thought was safe, including mathematics, law and finance.
The job you were counting on
You did what you were told. You got the qualification, took the entry-level role, put up with the boring parts because the boring parts were how everyone started. The plan was simple: a few years of learning, then a promotion, then the kind of salary that pays the rent without a knot in your stomach and leaves something over for your parents or your kids.
Now the boring parts are the first thing your manager hands to a tool. The junior posts on your team were not refilled when people left. A friend in customer support was told her team would be "restructured". You are still employed, and you are good at what you do, but you keep doing the sum: if a machine does the first three years of this job, how does anyone get to year four?
What the best data says about which jobs are exposed
The strongest evidence on which jobs AI touches comes from studying what people actually use it for, rather than guessing.
Microsoft Research analysed about 200,000 anonymised conversations with Bing Copilot and matched them to the work activities of every US occupation. The paper, first published in July 2025, found that AI is most often used for information work: gathering, writing and explaining. It gave each occupation an AI applicability score. The full scores are public. The highest-scoring jobs include interpreters, historians, writers, sales representatives, customer service staff and mathematicians. The lowest are almost all physical: dredge operators, roofers, dishwashers, orderlies.
The table below puts those scores next to the US Bureau of Labor Statistics projections for 2025 to 2035, released on August 27, 2026. All US jobs together are projected to grow about 3% over that decade.
| Occupation | AI applicability score (rank of 785) | US jobs 2025 | Projected change 2025 to 2035 |
|---|
| Interpreters and translators | 0.49 (#1) | 73,900 | +2% |
| Writers and authors | 0.45 (#3) | 140,300 | 0% |
| Customer service representatives | 0.41 (#7) | 2,666,000 | -5% |
| Telemarketers | 0.40 (#8) | 60,900 | -21.4% |
| Mathematicians | 0.39 (#10) | 2,200 | +1% |
| Data scientists | 0.36 (#19) | 275,600 | +35% |
| Personal financial advisors | 0.36 (#20) | 299,400 | +1% |
| Data entry keyers | 0.32 (#70) | 131,800 | -25.5% |
| Computer programmers | 0.31 (#81) | 110,800 | -7% |
| Software developers | 0.28 (#120) | 1,717,800 | +10% |
| Secretaries and administrative assistants | 0.26 (#143) | 3,515,600 | -2% |
| Bookkeeping, accounting and auditing clerks | 0.24 (#170) | 1,532,400 | -6% |
| Accountants and auditors | 0.19 (#255) | 1,595,200 | +5% |
| High school teachers | 0.18 (#299) | 1,087,500 | 0% |
| Registered nurses | 0.12 (#460) | 3,465,400 | +6% |
| Roofers | 0.01 (#770) | 166,900 | +5% |
Scores and ranks are from Microsoft Research's working-with-ai data. A higher score means more of the job's work activities overlap with tasks people successfully do with AI. It measures overlap with AI use. Whether a job shrinks depends on what employers do with that overlap.
Read down the projection column and a pattern appears. Where the whole job is a single information task (typing, data entry, a sales script, routine code), BLS already projects real decline. Where the job bundles information work with judgement, trust or a physical presence, BLS still expects growth, even when the AI score is high.
The top of the list shows the limits of the score. Interpreters and translators rank first of 785, yet BLS still projects 2% growth, and writers, ranked third, are projected to hold steady at 0%. A high score tells you the work inside the job is changing. Fewer jobs only follow when employers decide to hire fewer people, which is exactly what the entry-level data below shows.
The fastest-declining occupations in the BLS table for 2025 to 2035 read like a list of tasks AI does well: word processors and typists (-34.4%), telephone operators (-27.6%), switchboard operators (-26.0%), data entry keyers (-25.5%), telemarketers (-21.4%), order clerks (-17.5%), payroll and timekeeping clerks (-15.9%) and file clerks (-15.8%). For office and administrative support as a whole, BLS projects a fall of 4.0%, or 752,100 jobs, and says automation tools, "including those powered by AI," are likely to reduce demand.
Question to sit with: if you wrote down the five tasks that fill most of your week, how many of them are pure information work that a tool could draft?
Who is losing hours and entry-level roles now
Forecasts are about 2035. The clearest damage today is at the bottom of the ladder.
The Stanford Digital Economy Lab tracks millions of US workers through ADP payroll data. Its August 2026 paper by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen finds "no evidence of widespread, economy-wide job displacement." It also finds that employment of workers aged 22 to 25 in AI-exposed occupations "now stands 19% below where it would be had it kept pace with that of their less-exposed peers; experienced workers show no comparable gap." The gap has widened since the authors first reported it in August 2025. It works mainly "through reduced hiring of young workers rather than increased separations," and it is concentrated in occupations where AI substitutes for human tasks. Where AI mainly helps workers, employment is flat or rising.
That is the shape of the problem. Companies mostly keep the experienced people and stop hiring new ones.
Goldman Sachs Research found the same thing from another angle. In August 2025 it reported that unemployment among 20- to 30-year-olds in tech-exposed occupations had "risen by almost 3 percentage points since the start of 2025," more than for workers the same age in other trades.
The people using AI most see it too. Anthropic's June 2026 Economic Index report surveyed about 9,700 Claude users. Only 10% rated losing their own job in the next year as likely, but more than a third thought a junior colleague had a better-than-60% chance of losing theirs. Early-career respondents said AI could do the highest share of their work and were the most worried about job loss.
For new graduates, the New York Fed puts the unemployment rate for recent college graduates at about 5.6% in the second quarter of 2026, with 42% underemployed, meaning they work in jobs that do not need a degree. In its table by major, computer science graduates sit at 7.0% unemployment and computer engineering at 7.8%, among the highest of any field.
Question to sit with: if your firm stopped hiring juniors two years ago, who is going to be training under you, and who will need you to train them?
What companies are saying out loud
For years executives avoided linking layoffs to AI. That changed.
Challenger, Gray & Christmas, the outplacement firm that counts announced US job cuts, began tracking AI as a reason in 2023. In all of 2025, AI was cited for 54,836 cuts. By the end of September 2026, the year-to-date total was 120,136, about 21% of all announced cuts, making AI the leading reason of the year. In May 2026 alone it accounted for 40% of announced cuts. Technology firms account for 29% of all job cuts announced in 2026.
Two quotes show how plainly some leaders now talk. Salesforce chief executive Marc Benioff said on a podcast in late August 2025, about his support organisation: "I've reduced it from 9,000 heads to about 5,000 because I need less heads." Amazon chief executive Andy Jassy wrote to staff in June 2025 that generative AI gains mean "in the next few years, we expect that this will reduce our total corporate workforce."
Dario Amodei, chief executive of Anthropic, told Axios in May 2025 that AI could wipe out half of all entry-level white-collar jobs and push unemployment to 10% to 20% within one to five years. Critics noted he gave no research behind the number. It is a warning from someone who sells the technology, and you can weigh it as you like.
The jobs people thought were safe
The surprise of the last two years is how far up the skill ladder the exposure goes.
Mathematicians rank 10th of 785 occupations in the Microsoft data. In July 2026, AI models reported full marks at the International Mathematical Olympiad under official judging, where only 7 of 666 human contestants managed the same. Personal financial advisors rank 20th. Data scientists rank 19th, even as BLS projects their numbers to grow 35%. Software developers are further down at 120th, but computer programmers, the narrower coding job, are projected to shrink 7%.
Each of these has its own story, and this series covers them one by one:
Question to sit with: is your sense of safety built on what your job title sounded like ten years ago, or on what you actually do each day now?
What the official forecasts say, and where they disagree
The big institutional forecasts agree that AI will touch most jobs. They disagree sharply on how many people end up worse off.
The forecasts differ for clear reasons. The IMF and WEF count exposure and churn worldwide, so their numbers are large in both directions. Goldman models displacement in the US and notes that if only today's AI uses spread across the economy, about 2.5% of US employment would be at risk. BLS projects job totals from past trends, and it says its new AI exposure categories are "not a forecast of employment growth or decline." Stanford measures what is happening right now, and finds that the harm is real but narrow: young people, exposed jobs, fewer hires.
The honest counter-evidence matters. Total US employment is still projected to grow. BLS expects data scientists up 35%, registered nurses up 6% and roofers up 5%. Challenger's own September 2026 report shows announced cuts through September down 39% from the same months of 2025. AI is the leading stated reason for the cuts that do happen, in a year with fewer cuts overall.
None of that helps much if you are the 24-year-old whose application went into a system that no longer needs a junior. The averages are fine. The first rung is what is breaking.
What this means for you this month
You cannot control the forecasts. You can control how exposed your own income is.
Start with the table above. Find your job, or the closest one, and list the tasks you do that are pure information work. Those are the tasks to move away from or to own, by becoming the person who runs the tools instead of the person they replace.
Then build some income that does not depend on one employer's hiring plans. If you already use AI tools well, small businesses will pay you to set them up, which is the idea behind an AI automation agency. If you have a skill people struggle to learn, online tutoring pays by the hour from the first week. AI labs also pay people with degrees to test and grade their models, and this guide to AI training jobs covers the rates.
If you are already thinking about leaving, read how much money you need before you quit first, and if you are mid-career, changing careers at 40 covers how people have done it without starting from zero.
The rest of the series covers more jobs: engineers, architects, nurses, pharmacists, therapists, real estate agents, data analysts and graphic designers. For the forecasts side by side, see how many jobs AI will replace, and for the work least exposed, the AI-proof jobs post.