What Jobs Are NOT Safe From AI? (The Inverse View)
Our pillar guide covers what jobs are safe from AI. This is the flip side: what jobs are not safe from AI, and why.
The short answer is that the exposed work is routine, repeatable, and low-stakes if it's wrong. Data entry, first-draft writing, basic customer support scripts, standard bookkeeping, junior document review, and entry-level coding tasks are the ones current AI already does at or near human level, and the labor-market data backs that up.
This isn't a doom list. Being in an exposed job doesn't mean losing it tomorrow, and every field below still has roles inside it that are holding up fine. But naming the exposed work honestly is more useful than pretending everyone's fine, so let's get specific.
Why these jobs are the ones getting automated
Microsoft Research's 2025 study, built from 200,000 real conversations with the company's Copilot assistant, mapped what people actually use AI to do against the government's occupational database. You can read the full paper directly. The researchers found the highest overlap between real AI use and job tasks concentrated in roles built around producing text, processing information, and answering routine questions, exactly the tasks a large language model is designed to do well.
The researchers were careful to note, in a follow-up post, that high applicability measures overlap with tasks, not certainty of job loss. But when you combine that overlap data with actual government employment projections, several categories show a consistent pattern: shrinking headcount alongside rising automation, not just theoretical exposure.
The jobs and tasks most exposed right now
Routine data entry and bookkeeping
The clearest case. The U.S. Bureau of Labor Statistics projects employment of bookkeeping, accounting, and auditing clerks to decline 6 percent from 2024 to 2034, citing software automation directly as the cause. The BLS is explicit: the same amount of work can now be done with fewer people, because software handles the routine entry and reconciliation that used to require a dedicated clerk. Despite the decline, the BLS still projects about 170,000 openings a year from people leaving the field, so the job doesn't vanish overnight, but the long-term headcount trend is down.
Standard customer support
Customer service representative employment is projected to decline 5 percent from 2024 to 2034, per the BLS, with the agency pointing to self-service systems, chatbots, and mobile apps letting customers handle simple requests without a person. The exposed part of this job is specifically the routine, scripted interaction: password resets, order status, basic troubleshooting. Complex complaints, emotionally charged situations, and anything requiring a judgment call about an exception to policy are a different story, and those tasks are exactly where human reps are still needed.
First-draft writing and routine content
Any job built mainly around producing a first version of standard text (boilerplate reports, templated emails, routine marketing copy, basic summaries) sits inside the highest-overlap zone in the Microsoft data. This isn't a single job title so much as a task that shows up across many roles. The people most exposed are the ones whose value was mostly "I produce the first draft," rather than the ones who edit, direct, or take responsibility for what gets published.
Junior-level document review
Paralegal and legal assistant employment is projected to show little or no change from 2024 to 2034, and the BLS specifically names AI as one of the technologies limiting growth in this field. The task most affected is first-pass document review: skimming large volumes of material for keywords, patterns, or relevant clauses, which overlaps heavily with what generative AI already does well. Senior paralegal work involving client contact, case strategy, and courtroom support is a different, less exposed layer of the same job title.
Entry-level coding and basic scripting
Writing routine, well-specified code (boilerplate functions, standard scripts, simple bug fixes) is squarely inside current AI capability, and it's the task most junior developers historically cut their teeth on. This doesn't mean software engineering as a field is unsafe. It means the traditional entry ramp, where junior developers built skill by doing simple, repetitive coding tasks under supervision, is changing shape, and the more durable part of the job (architecture decisions, debugging genuinely novel problems, understanding what the business actually needs) sits higher up the same career ladder.
What these jobs have in common
Every job and task on this list shares two traits. First, the work is repeatable: today's task looks a lot like yesterday's, with enough consistency that a model trained on patterns can produce a usable result. Second, the stakes of a single mistake are low enough that organizations are comfortable letting AI handle the first pass, with a human checking only the exceptions.
Compare that to the categories in our pillar guide on what jobs are safe from AI: physical, place-specific trades and high-accountability human-judgment roles. Those hold up precisely because they lack both traits: the work isn't repeatable in the same way, and the cost of a mistake is too high to hand to a system that can be confidently wrong.
The World Economic Forum's Future of Jobs Report 2025 frames this at the skills level: employers rank analytical thinking, resilience, and creative thinking as the most in-demand core skills, and estimate 39 percent of workers' core skills will shift by 2030. Read plainly, that's a description of tasks moving away from routine output and toward judgment, inside the very jobs on this list, not a prediction that the jobs disappear.
What to actually do if your job is on this list
Being in an exposed role is a signal to act, not a verdict. A few grounded moves:
Move up the same ladder. If your job includes both the routine task and a layer of judgment above it (checking the work, deciding what matters, handling exceptions), lean hard into that layer. It's the part staying valuable.
Get fluent with the tools instead of competing against them. People using AI to produce more, faster, tend to be pulling ahead of people avoiding it. The task getting automated is an opportunity to redirect your time, not just a threat.
Watch the trend, not just the snapshot. A 5 or 6 percent decline over a decade isn't a cliff. It's a slow shift that gives you time to build the skills that sit above the exposed task, if you start now rather than later.
If you're specifically trying to plan a move, our companion piece on what jobs is AI replacing looks at the same question from the industry-trend side, and it's worth reading alongside this one for the fuller picture.
Find out exactly where your job stands
General categories are a starting point, but your actual exposure depends on your specific daily tasks, not just your job title. Two people with the same title can be in very different positions.
Take the free "How AI-Proof Is Your Job?" assessment to get a personal read in a few minutes and see what people in similar roles are doing about it.
Five days to take back your core tasks
The tasks you still do by hand, without checking a tool first, are the ones that keep you valuable. 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.
Frequently asked questions
What jobs are most at risk from AI?
Roles built mainly around routine, repeatable tasks with low individual stakes: data entry, standard bookkeeping, scripted customer support, first-draft writing, junior document review, and entry-level coding. BLS data shows real employment declines in several of these categories already underway through 2034.
Will AI replace customer service jobs?
The routine parts are already being automated. BLS projects customer service representative employment to decline 5 percent from 2024 to 2034, driven by self-service tools and chatbots handling simple requests. Complex, judgment-heavy, or emotionally sensitive interactions still need a person.
Are entry-level jobs disappearing because of AI?
Some entry-level tasks are shrinking, particularly ones built around producing a first draft or doing routine data processing. That's changing how people build skill early in a career more than it's eliminating entire fields. The senior roles in the same fields, built on judgment and oversight, are generally holding up.
Is it too late to switch careers if my job is exposed?
No. A gradual decline over a decade, which is what most BLS projections here actually show, gives real time to build skills that sit higher on the same ladder or to move toward a less exposed field. The workers struggling most are the ones who wait rather than the ones who start adjusting now.
How do I know if my specific job is at risk?
Look at how much of your day is spent on repeatable tasks with low individual stakes versus judgment calls and exceptions. Our free assessment scores this directly using your actual tasks rather than your job title, and shows how your result compares to similar roles.
Knowing which jobs are exposed only matters if you know where you personally stand. Take the free "How AI-Proof Is Your Job?" assessment to see your real score and what to do next.