Jobs least affected by AI, and what makes them hold up

You work in customer service for a mid-size company, and this week two people on your team got a termination email the same day the company announced a new AI chat tool. Nobody said the two things were connected. Everyone on the floor knows they were. You pull out your phone that night and type "jobs least affected by AI" into the search bar, half hoping for a tidy list, half expecting nothing useful. There is a useful answer, and it has less to do with which industry you pick and more to do with what the job actually asks you to do all day.

What the AI impact score actually measures

An AI impact score, rated out of 10 with higher meaning more exposure, is a rough way of asking one question: how much of this job is repeatable text, numbers or patterns that a model can copy from examples. Writing a summary, drafting a contract clause, sorting an inbox, flagging a claim: these are tasks that score high because they're made of language and data, the exact material AI tools are trained on. A job scores low when most of the work happens through a body in a room with another body, or through judgment calls made on the spot, with consequences that matter immediately. A firefighter decides which door to force open. A hairstylist reads a client's face in the mirror mid-cut. A toddler who won't nap needs someone to notice why, right now, not after a model finishes processing the request. None of that fits into a training set the way an email thread does.

The jobs with the lowest AI impact score

Here's how the professions with the lowest AI impact scores in this comparison stack up against each other on pay and demand.

A few things jump out once you sit with this table instead of just skimming it. The five jobs with a score of 1, recreation worker, preschool teacher, childcare worker, firefighter and hairstylist, pay less on average than the four jobs sitting at a score of 2. Physical therapist alone earns $102,760 a year, close to three times what a childcare worker makes at $34,980. Low AI exposure and high pay are not the same thing, and this table is the proof. The second thing worth noticing is that a low score doesn't guarantee a growing field. Childcare worker has the lowest job outlook in the whole group, at a score of 2 for decline, even though no algorithm is coming for that work. The threat there isn't automation. It's a staffing and funding problem, which calls for a different kind of fix. Meanwhile hairstylist and physical therapist both sit at a job outlook of 9, much faster than average, which means a low AI score and strong growth can line up, just not automatically. Openings per year swing just as widely, from 13,400 for physical therapist, a role that requires a doctorate, to 196,200 for certified nursing assistant, a role you can enter with 75 or more hours of training plus a state exam.

Why these jobs resist automation

Look at what the five score-1 jobs have in common and a pattern shows up fast. Every one of them puts a person's safety or well-being directly in someone else's hands, in real time, with no undo button. A firefighter can't pause mid-rescue to check a model's suggestion. A preschool teacher reading a room of four-year-olds is making dozens of small judgment calls a minute, about who's about to cry and who needs a different activity entirely, decisions that don't repeat in any pattern clean enough to automate. A hairstylist is working with a different head, a different hair texture and a different mood every forty-five minutes, which is part of why the job keeps a 9 out of 10 outlook score even as other service jobs flatten out. Recreation workers sit in the same bucket for a related reason: the job is built around reading a group's energy and adjusting a plan on the fly, something that lines up with the work-life balance score of 7 in this table, the highest of any job listed, because so much of the work happens in scheduled blocks of activity rather than an inbox that never empties.

Move one step up to the score-2 jobs and the pattern bends without breaking. A landscaper, a certified nursing assistant, a physical therapist and a carpenter all still work with their hands on a physical thing or a physical person, but a slice of each job, scheduling, measurements, documentation, can be nudged along by software. That's the real difference between a 1 and a 2 in this data: not whether AI touches the job at all, but how much of the job is left standing once it does.

What you give up for that safety

None of this is free. Look at the work hours listed for firefighter: full time, often on 24-hour shifts followed by 48 hours off, and overtime is common. That schedule protects the job from a chatbot, but it also means missed dinners and holidays on a rotation you don't choose. Certified nursing assistant carries a work-life balance score of 4, the lowest in the table, tied to a reality where nights, weekends and holidays are part of the job because hospitals and nursing homes run around the clock. Childcare worker and preschool teacher both sit at a work-life balance score of 5, and preschool teacher's education route still requires an associate or bachelor's degree, two to four years of training, for a starting salary of $28,990 on the low end of its range.

The training question matters just as much as the schedule. Hairstylist asks for about 9 to 15 months of training, short compared to a four-year degree, and openings run at 73,400 a year with a job outlook of 9 out of 10, which makes it one of the fastest routes into a role that's genuinely hard to automate. Physical therapist sits at the other extreme: three years of doctorate-level training after a bachelor's degree, for a salary ceiling of $135,140 on the high end, the best pay in this entire list. Plumber and carpenter, both sitting at a score of 2, land in between, each asking for several years of apprenticeship, four to five years for plumber and about four years for carpenter, in exchange for salaries that top out at $108,420 and $99,910. These are not short detours. Weigh the training length against your current runway, not against the headline salary, because the gap between applying and earning full pay is where most people underestimate the cost of switching.

The schedule you're actually signing up for

Pay and training length aren't the only trade-offs hidden in this list. Work hours vary enormously from one AI-resistant job to the next, and ignoring that is how people end up leaving a "safe" job within a year. Firefighter runs on 24-hour shifts followed by 48 hours off, which sounds generous until you factor in the overtime that's described as common in the field, meaning the schedule looks like time off on paper but rarely feels that way in practice. Certified nursing assistant work covers nights, weekends and holidays because hospitals and nursing homes operate around the clock, and shift length varies by employer, so two CNAs at different facilities can end up with completely different weeks. Recreation worker sits at the other end of the spectrum: many workers in that field are part time or seasonal, concentrated in summer, with schedules built around afternoons, evenings and weekends, exactly when the public has free time to show up for programs. Hairstylist offers something closer to control than any other job on this list, since many stylists set their own hours, especially once they rent a booth or go self-employed, though evening and weekend work stays common because that's when clients want appointments. Physical therapist lands in the middle: mostly full time, with scheduled appointments that make the week predictable, aside from occasional extra hours spent on documentation after the last patient leaves. None of these schedules showed up in the AI impact score, but all of them decide whether the job is livable once you're actually doing it.

Moving from a score of 2 toward a score of 1

If you're currently in a role with real AI exposure and you're eyeing something on this list, the honest move is to pick based on what you already do well, not based on which score looks smallest. Someone who's spent years managing a chaotic retail floor already has the pattern-reading instinct that recreation worker and preschool teacher both lean on, just aimed at kids or program participants instead of customers. Someone who already works well with their hands and doesn't mind repetitive, physical tasks might find landscaper's path shorter than it looks: the education route is on-the-job training, a license or a certificate, with training taking a month or less, the fastest entry point in this whole comparison, though it comes with a work-life balance score of 4 and 61% of workers logging more than 40 hours a week during peak season. Someone with steadier mechanical instincts, the kind that enjoys fixing what's broken rather than planning around a group's mood, might look at plumber instead, which asks for four to five years of apprenticeship but pays up to $108,420 on the high end.

School counselor is worth a second look for anyone coming from teaching or school-based social work, since it asks for a bachelor's and master's degree but offers a work-life balance score of 7, tied with recreation worker for the best in the group, built around a school-day schedule with summers off for many positions. Line cook sits at a score of 2 with a striking growth number: a 12.1% expected job growth rate and 218,600 openings a year, the highest opening count in the entire dataset, which makes it one of the easiest fields to get a foot into even though the pay tops out modestly at $47,900.

If you're starting from zero

If none of your current skills point toward any of these yet, start smaller than a full career change. Pick one job from the table that pays enough to cover your bills during training, and check the shortest education route listed for it before you check anything else. That number decides whether this is a six-month plan or a four-year one. Line cook is worth starting with for exactly that reason: the training is months on the job, not years in a classroom, openings run at 218,600 a year, the highest count in this whole comparison, and the AI impact score sits at just 2 out of 10 because no model can taste a sauce or time a dozen tickets firing at once during a dinner rush. The pay tops out modestly, at $47,900 on the high end of the range, but it's a real foothold, not a stopgap. Look at the Line Cook page this week, read what a typical kitchen expects from someone with zero experience, and ask one restaurant near you if they're hiring for a line position right now.