Which Jobs Will AI Replace First?
The jobs AI is coming for first aren't the ones sci-fi warned about. They're not truck drivers or factory workers: they're desk jobs made of routine language and data.
The jobs AI is coming for first aren't the ones sci-fi warned about. They're not truck drivers or factory workers: they're desk jobs made of routine language and data. Interpreters, translators, writers, customer service reps, and sales support are at the front of the line, because their work is exactly what today's AI does well: reading, summarizing, drafting, answering. If your day is mostly words and screens, you're closer to the front than someone who works with their hands.
But "which jobs will AI replace" is the wrong question if you take it literally, and getting it slightly wrong will either scare you for no reason or leave you complacent. AI rarely replaces a whole job. It replaces tasks, and jobs made almost entirely of the exposed tasks are the ones that shrink. So the real list isn't "doomed occupations." It's "occupations where the automatable tasks are the whole job." Let's go through the evidence and build that list honestly.
What the exposure data shows
The most detailed read comes from Microsoft Research, which analyzed 200,000 real conversations with Copilot to measure how much each occupation's tasks overlap with what AI can already do, scoring 785 occupations by how often AI is used for their work and how well it performs (Microsoft Research).
The top of the list, ranked by AI applicability (Fortune):
- Interpreters and translators (the single highest score)
- Historians
- Writers and authors
- Customer service representatives
- Sales representatives of services
- Telephone operators and telemarketers
- Broadcast announcers and radio DJs
- Ticket agents and travel clerks
- Proofreaders and copy markers
- Technical writers, editors, and PR specialists
Notice the pattern. Every one of these is knowledge work built on producing or communicating language and information. Interpreters and translators scored highest because their tasks show up constantly in AI conversations and the AI completes them well (Visual Capitalist).
At the bottom of the exposure list: roofers, logging and equipment operators, and other physical, hands-on roles (GeekWire). Machines that operate in the messy physical world are still hard and expensive. Software that manipulates text is cheap and everywhere.
One crucial caveat, straight from the researchers: this measures exposure, not elimination. Microsoft was direct that a high score means AI touches the tasks, not that the job disappears, and that most roles are more likely to be reshaped than removed (Microsoft Research blog). Hold onto that. Exposure is where to look first. It's not a verdict.
The other pattern: it hits the bottom rung first
Exposure tells you which occupations. A Stanford study tells you which people within them get hit first, and the answer is uncomfortable.
Economist Erik Brynjolfsson and colleagues analyzed ADP payroll data covering 25 million workers. Early-career workers aged 22 to 25 in the most AI-exposed jobs saw a 13% relative decline in employment after generative AI arrived, while older workers in the same occupations stayed flat or grew (Fortune). In entry-level software engineering and customer service specifically, roles fell by nearly 20% between late 2022 and mid-2025 (TIME).
Why the young first? Because the tasks juniors are given, the routine, learnable, repetitive slice, are the most automatable part of any exposed job. AI didn't come for the senior analyst's judgment. It came for the grunt work that used to train the next senior analyst. That's the sharpest version of "which jobs will AI replace": the entry-level tier of exposed white-collar work is replacing first.
Brynjolfsson also drew the line that matters most: the damage concentrates where AI automates tasks rather than augments them (CBS News). Same technology, opposite outcomes, depending on whether it does your work or makes you faster at it.
The occupations most exposed, and why
Pulling the evidence together, here's the honest first-wave list, with the reason each one's on it.
Customer service and support. High-volume, scriptable, digital. AI handles routine tickets well, which is why customer service shows up near the top of every serious analysis and in the Stanford entry-level numbers. About 5 million US jobs sit in this category (Fortune).
Translation and interpretation. Machine translation crossed a quality threshold that makes it good enough for a large share of routine work. The human edge is now in high-stakes, legal, literary, and live-negotiation contexts.
Writing and editing of routine content. Product descriptions, summaries, templated articles, first-draft copy: squarely in AI's wheelhouse. Investigative, deeply reported, or highly original writing is a different story; the label "writer" hides that gap.
Data entry, clerical, and administrative support. Structured, digital, repetitive. This is the category the WEF flags for the largest routine-role losses.
Entry-level software development. Boilerplate coding and simple bug fixes are increasingly AI-assisted, which is why junior roles took a visible hit while senior, architecture-level work held up.
Telemarketing, phone operators, and basic sales support. Scripted outbound and inbound calling maps cleanly onto what AI voice systems now do.
Compare that to the roles the same research puts at low exposure: skilled trades, hands-on healthcare, equipment operation, and anything requiring physical presence or licensed accountability. If you want the mirror image of this list, the roles best positioned, see The Most AI-Proof Jobs.
Which jobs won't be replaced (and why the line isn't the title)
It's tempting to end with "so avoid the exposed jobs." That's too crude, because exposure lives at the task level, not the title level.
A customer service role that's pure ticket-closing is highly exposed. A customer service role that's really account management (retaining angry enterprise clients, negotiating, owning relationships) is not, even under the same job title. A writer cranking out SEO filler is exposed; a writer whose value is original reporting and a real byline is not. The occupation label is on the list. The specific person may or may not be.
This is why we push a task audit over a title lookup. To see whether your version of an exposed job is actually at risk, run the free How AI-Proof Is Your Job? assessment: it scores your real tasks, not your occupation's average, and shows how people in similar roles are repositioning. And for a role-by-role walkthrough of eight common jobs, our companion piece Is Your Job At Risk From AI? goes deeper on specific titles.
What to do if your job is on the list
Being on an exposure list is a prompt to reposition, not a reason to quit. The move is the same regardless of role: shift your time from the automatable slice toward the parts of your work that need judgment, relationships, physical presence, or accountability, and get good at directing AI so you're the operator, not the task being replaced.
Keep the aggregate in view so the fear stays proportionate. The World Economic Forum projects about 92 million jobs displaced by 2030 but roughly 170 million created (a net gain of 78 million), with heavy churn along the way (World Economic Forum). The jobs are moving, not vanishing. The people who come out ahead are the ones who see the move coming and step, rather than waiting to be pushed.
Frequently asked questions
Which jobs will AI replace first?
The routine, language-and-data desk jobs: customer service, translation and interpretation, clerical and data-entry work, routine writing, entry-level coding, and telemarketing. Microsoft's research ranks these highest for AI exposure (Fortune). Physical and hands-on jobs are least exposed.
Will AI replace entire jobs or just tasks?
Mostly tasks. Whole-job replacement happens only where a job is made almost entirely of automatable tasks. Microsoft's researchers stressed their scores measure exposure, not elimination (Microsoft Research blog).
Why are entry-level jobs hit hardest?
Because junior tasks are the most routine and learnable part of a job: exactly what AI automates first. Stanford found a 13% relative employment drop for 22-to-25-year-olds in the most exposed occupations, while older workers held steady (Fortune).
Which jobs are safest from AI?
Skilled trades, hands-on healthcare, and roles built on physical presence, complex judgment, or licensed accountability sit lowest on the exposure list. See The Most AI-Proof Jobs for the full picture.
How do I know if my specific job is at risk?
Audit your tasks, not your title. Score each task on whether AI can do it now and whether a human has to be accountable. The free How AI-Proof Is Your Job? assessment does this and compares you to similar roles.
Want a straight answer for your own role? Take the free, anonymous How AI-Proof Is Your Job? assessment and find out where you actually stand.
Published August 2026.