AI is far more likely to replace specific tasks than entire jobs, and the roles at highest risk share one trait: they're built on routine, predictable work that a machine can learn. The clearest targets right now are data entry clerks, bank tellers, basic bookkeepers, and parts of office and administrative support. The roles holding up best involve physical dexterity, hands-on judgment, or human relationships that are hard to script.
Here's the short version before we dig in:
- The World Economic Forum projects 92 million jobs displaced by 2030, but 170 million new ones created in the same window.
- Office and administrative support has the highest share of automatable tasks, with Goldman Sachs estimating up to 46% of those tasks could be done by AI.
- Roughly 19% of U.S. workers are in jobs most exposed to AI, while 23% are in the least exposed, according to Pew Research.
- "Exposed" rarely means "eliminated." Most affected jobs change shape rather than disappear.
The honest answer to "what jobs will be replaced by AI" is: fewer than the headlines suggest, but the disruption to *how* work gets done is real. Below is what the data actually says.
How to Tell If a Job Is at Risk
A job's exposure to AI comes down to what the work is made of, not the job title on a business card. The more a role depends on repetitive, rule-based tasks with clear inputs and outputs, the easier it is to automate.
Pew Research defines exposure as the likelihood that the most important activities in a job could be replaced or assisted by AI. Using O*NET work-activity data, the analysis found that 19% of American workers held jobs in the most exposed group, where core activities sit squarely within reach of current AI. Education tracks closely with exposure: only 7% of workers with a bachelor's degree or higher were in the least-exposed group, compared with half of workers without a high school diploma.
Three signals tend to predict high risk:
- Routine and repetition. Tasks that follow the same steps every time, like sorting records or processing standard forms.
- Digital, text-heavy work. Drafting boilerplate, summarizing documents, and basic data lookup are exactly what large language models do well.
- Low need for physical presence or improvisation. A job done entirely at a screen is easier to automate than one requiring hands, mobility, or on-the-spot human judgment.
Jobs that score low on all three, such as electricians, nurses, and skilled trades, are far harder to hand to a machine.
The Jobs Most Likely to Be Replaced
Several major studies point at the same cluster of roles. The World Economic Forum's Future of Jobs Report 2025 names clerical and secretarial work as the fastest-declining category in absolute terms, with cashiers, administrative assistants, postal clerks, bank tellers, and data entry clerks at the front of the line.
Goldman Sachs estimated that generative AI could expose the equivalent of 300 million full-time jobs to some automation, with office and administrative support showing the highest share of automatable tasks at 46%, followed by legal work at 44% and architecture and engineering at 37%. As CNBC's breakdown of the Goldman report noted, administrative staff and lawyers are expected to feel the largest effects, while physically demanding outdoor work like construction sees little.
Here's how the highest-risk roles stack up:
| Role | Why it's exposed | What AI replaces |
|---|---|---|
| Data entry clerk | Repetitive, structured input | Most of the core task |
| Bank teller | Transactions already shifting to apps and kiosks | Routine handling |
| Bookkeeper / basic accounting | Rule-based number work | Reconciliation, categorization |
| Telemarketer / customer service rep (tier 1) | Scripted conversations | Standard inquiries |
| Proofreader / basic copy editor | Pattern-based text checking | First-pass review |
| Paralegal (document-heavy tasks) | Searchable, templated documents | Discovery, drafting |
A key caveat runs through all of this: even in highly exposed occupations, Goldman estimated that only a quarter to roughly half of the workload could actually be automated. That usually reshapes the job rather than erasing it.
The Jobs AI Is Least Likely to Replace
The roles holding up best aren't necessarily high-tech. They're the ones built on things AI still can't do well: physical work in unpredictable environments, genuine human care, and complex judgment under ambiguity.
Skilled trades top the list. Electricians, plumbers, HVAC technicians, and mechanics work in messy, varied physical settings where every job is a little different, which is the opposite of what automation thrives on. Healthcare roles that involve hands-on care and human trust, like nurses, physical therapists, and home health aides, are similarly resistant. So is anything that hinges on persuasion, negotiation, and relationship-building, which is why interpersonal skills like coaching and negotiation are projected to change the least.
If you want a fuller breakdown of durable roles, our guide to AI-proof jobs covers specific occupations and the skills behind them. For anyone weighing a move into tech, the IT career path guide walks through which roles are growing fastest.
A few categories standing on firm ground:
- Skilled manual trades requiring dexterity and on-site problem solving.
- Care and frontline healthcare built on human contact.
- Complex strategy, leadership, and creative direction that set goals rather than execute set tasks.
- Frontline growth roles the WEF flags as expanding, including delivery drivers, construction workers, and salespeople.
What This Means for Your Career
The reassuring part of the data is the net number: the WEF expects more jobs created than destroyed by 2030, and history backs the pattern that displacement is usually followed by new categories of work. The uncomfortable part is the churn. On average, workers can expect roughly 39% of their current skill set to be transformed or outdated over the 2025-2030 stretch.
That points to a clear move: get good at using AI rather than competing with it. Pew found that about 21% of U.S. workers already use AI in their jobs, and that share is climbing. The workers least at risk in any field are the ones who use these tools to do more, faster.
Three practical steps:
- Shift toward the parts of your job AI can't do. Judgment, relationships, physical skill, and creative direction.
- Learn the tools in your field. A bookkeeper who masters AI reconciliation software is more valuable, not less.
- Build a real network before you need one. When roles shift, the people who land well are the ones with direct relationships, not the ones starting cold. Our guide on how to get a job goes deeper on this.
That last point matters more as AI floods the application funnel. When résumés can be generated in seconds and screened by algorithms, the candidates who stand out are the ones who reach a hiring manager directly. If you're planning your next move in an AI-shaped market, Articuler helps you find the actual person hiring behind a posting using semantic search across 980M+ profiles, then surfaces the right people to reach and drafts outreach that earns a reply instead of disappearing into another ATS. The fastest path through a changing job market is still a real conversation with the right person.
FAQ
Will AI replace most jobs by 2030? No. The World Economic Forum projects 92 million jobs displaced but 170 million created by 2030, a net gain. AI is far more likely to automate specific tasks within jobs than to eliminate entire occupations.
Which jobs are most at risk from AI? Routine, screen-based roles face the highest risk: data entry clerks, bank tellers, basic bookkeepers, telemarketers, tier-one customer service reps, and document-heavy administrative work. Goldman Sachs found office and administrative support has the highest share of automatable tasks at 46%.
Which jobs are safest from AI? Skilled manual trades, hands-on healthcare, and roles built on negotiation, leadership, and creative judgment hold up best. Jobs requiring physical dexterity in unpredictable settings or genuine human trust are the hardest to automate.
Does "exposed to AI" mean the job will disappear? Not usually. Exposure means some of a job's core tasks could be automated or assisted by AI. Even in the most exposed occupations, Goldman Sachs estimated only a quarter to about half of the workload could be automated, which reshapes jobs more often than it erases them.
How can I protect my career from AI? Move toward tasks AI can't do, learn the AI tools in your field, and build direct professional relationships. Roughly 39% of current skill sets are expected to change by 2030, so adaptability and a real network matter most.
- /
- /product/find-the-right-people/