Jobs Most at Risk From AI (The Honest List)
Yes, some jobs are genuinely at high risk from AI right now, and the honest answer names them instead of softening it. The jobs most exposed are desk-based, language-heavy roles: translation, writing, customer service, and administrative work where the core output is text or spoken language a model can already produce.
This is the mirror image of jobs least likely to be replaced by AI, and the two lists are best read together, because the pattern that protects one group is exactly what's missing from this one: no physical presence requirement, no license, no direct legal liability standing between the task and the software doing it.
The occupations with the strongest evidence
Microsoft's 2025 study measuring generative AI's applicability across hundreds of occupations put interpreters and translators at the very top, followed by historians, passenger attendants, writers and authors, and customer service representatives, an occupation covering roughly 5 million U.S. jobs. Also on the top-40 list: telemarketers, proofreaders and copy markers, technical writers, news analysts and reporters, brokerage clerks, and public relations specialists.
The pattern across nearly the entire list: the job's output is language, produced from a desk, with no equipment to operate and no license required to do it. A translator converts text to text. A customer service rep answers a question using information that already exists somewhere. Those are close to exactly the tasks a large language model is built to do.
What BLS actually projects for these specific occupations
The Microsoft exposure score and the government's own hiring forecast tell two different parts of the same story, and both belong in view rather than picking whichever sounds worse. BLS projects customer service representative employment to decline 5% from 2024 to 2034, with about 341,700 openings a year anyway, almost all from people leaving the occupation rather than new hiring. BLS names the cause directly: continued automation of routine tasks, plus self-service tools letting customers skip the representative entirely.
Interpreters and translators, the single most exposed occupation on Microsoft's list, tell a stranger story. BLS projects that occupation to grow 2% from 2024 to 2034, slower than average but still growth, driven by continued demand from a more diverse population and by military and national-security translation needs that AI hasn't absorbed. The honest read: even the occupation with the highest AI-capability overlap in Microsoft's entire study is projected to grow, only more slowly than it otherwise would have. That's a real cost to the people who would have gotten those jobs, but it's a different claim than "translators are being replaced," and the difference matters if you're deciding whether to leave the field.
What's actually already happening
The risk also shows up in labor decisions already being made. Reporting cited in Fortune's coverage of the Microsoft study notes companies including Amazon and Meta announcing workforce reductions in parallel with heavy AI investment, and hiring pauses at firms like Norway's sovereign wealth fund specifically tied to AI productivity gains. That is a correlation. It does not prove that any single layoff was caused by AI rather than broader cost-cutting, and it should be read that way. But it means the risk is no longer purely theoretical.
BLS's own 2024-2034 employment projections name automation, including AI specifically, as a contributor to declining office and administrative support employment over the decade, alongside declining demand in production occupations from continued automation. That is a government projection, and it points the same direction.
The one nuance that matters
Microsoft's own researchers were explicit that a high applicability score does not automatically mean a job gets eliminated. It means the job's tasks closely overlap with what generative AI can currently do. A customer service representative whose job is 70% answering repeated questions and 30% de-escalating an angry, unusual case is exposed on the first part and not on the second. What tends to happen first is the job needing fewer people to do the same volume of work, with the survivors doing more of the 30% and less of the 70%.
That distinction is cold comfort if you're one of the people not needed anymore, and it's dishonest to pretend otherwise. Pew's 2025 research found 32% of U.S. workers now expect AI to mean fewer job opportunities for them personally over the long run. That worry tracks the real data reasonably well for the occupations on this list.
A real deployment, and its walk-back
Klarna's 2024 rollout is the clearest documented case of this risk playing out, including the part that usually gets left out. In February 2024, the company launched an OpenAI-built customer service assistant that handled 2.3 million chats in its first month, doing the equivalent work of 700 full-time agents and resolving issues in under two minutes versus 11 for a human, according to reporting on the deployment's first-year numbers. Klarna's own headcount fell from about 5,000 to 3,500 over the following year, largely through attrition rather than direct layoffs.
Then, in May 2025, Klarna's CEO told Bloomberg the company had cut too far and began rehiring human agents, after customers complained the AI gave generic answers and struggled with anything outside a scripted pattern. That reversal is the key detail: the technology handled real, measurable volume, but the company found the last mile, the unusual or emotionally loaded case, needed a person after all. That's a specific, cited example of the exact split described above: repeatable volume goes to the model, the harder remainder stays human, and a company can misjudge where that line actually sits.
What to do if your job is on this list
Two moves matter more than generic "learn AI" advice. First, get specific about which part of your job is the repeatable, templatable part versus the part that requires judgment on something unusual, a relationship, or a decision someone has to own. Second, move toward the second category deliberately, inside your current role if you can, rather than waiting to be pushed there. A customer service rep who becomes the escalation specialist is in a different position than one who stays a first-line responder.
The tasks you keep decide how replaceable you are
The tasks you can still do without leaning on AI are what make you hard to replace here. The free 5-Day AI Reset is a five-email course built around exactly that: Day 2 has you take one task back and do it unassisted. One small change per day, and it stays useful no matter how the job market moves.
Frequently asked questions
What jobs are most at risk from AI right now?
Interpreters and translators, historians, writers and authors, customer service representatives, and other desk-based, language-output roles score highest on Microsoft's 2025 occupational applicability study measuring overlap with generative AI's current capabilities.
Is customer service really at risk from AI?
Yes, for the repeatable-question portion of the job. Customer service representatives, covering about 5 million U.S. jobs, rank in the top 10 most-exposed occupations in Microsoft's study, though escalation and relationship-heavy parts of the role remain harder to automate.
Does high AI exposure mean a job will disappear?
Not necessarily and not immediately. Microsoft's researchers were explicit that high applicability measures task overlap with AI capability, not a prediction of elimination. BLS's 2024-2034 projections show automation reducing growth in office and administrative categories rather than eliminating them outright.
Are layoffs actually happening because of AI?
Some companies, including Amazon and Meta, have announced workforce reductions alongside heavy AI investment, and some firms have paused hiring citing AI productivity gains. Whether AI specifically caused any individual layoff versus broader cost-cutting is usually not verifiable from public statements alone.
What should I do if my job is on this list?
Identify which specific tasks in your role are repeatable and templatable versus which require judgment, relationship, or handling something unusual, and move toward the second category inside your current job rather than waiting for the decision to be made for you.
Do BLS projections show these jobs actually disappearing?
No, not for most of them. BLS projects customer service representatives to decline 5% and interpreters and translators to grow 2% from 2024 to 2034. That's a slowdown, in one case a real decline, but neither is a projection of the occupation vanishing.
Want to know where your specific role falls instead of guessing from a list? Take the How AI-Proof Is Your Job quiz.