Jobs AI won't replace: 60+ AI-proof careers and why they resist

The jobs AI won't replace any time soon share a few traits: they happen in the physical world, in unpredictable places, with people who need to trust another person, and with someone legally accountable for the result. Electricians, nurses, care workers, plumbers, mechanics, early years educators and many skilled trades sit firmly in that group. Below: what the latest research actually says, 60+ AI-proof careers by family, the most exposed jobs, and how to choose.

What the research says about jobs safe from AI

Since 2025, several serious studies have measured how AI is really used at work. Their methods differ, but they point in the same direction.

Anthropic (March 2026): real usage, not just theory

In Labor market impacts of AI: A new measure and early evidence, published on 5 March 2026, Anthropic's economists built an "observed exposure" score. It looks at the tasks a language model could theoretically speed up, then checks which of those tasks actually show up in professional use of Claude. Computer programmers came out on top, with about 75% of their tasks covered, followed by customer service representatives and data entry keyers (about 67%). At the other end, roughly 30% of workers had zero coverage because their tasks barely appeared in the data at all. The examples given are telling: cooks, motorcycle mechanics, lifeguards, bartenders, dishwashers and dressing room attendants.

Just as important is what the study did not find. The authors report no systematic rise in unemployment among highly exposed workers since late 2022. They do see "suggestive evidence" that hiring of 22 to 25-year-olds has slowed in exposed occupations, while noting that this result is only barely statistically significant and could have other explanations.

Microsoft (2025): where generative AI applies

Microsoft researchers analysed 200,000 anonymised conversations with Bing Copilot in the US (January to September 2024) to score how applicable AI is to each occupation. Information work ranked highest: media and communication workers, sales representatives for services, and information and record clerks. The lowest scores went to forest and conservation workers, grounds maintenance workers, plant and system operators, and home health aides and nursing assistants. Among large occupational groups, construction trades workers and building cleaning workers were also near the bottom. The authors warn explicitly that a high score does not mean job losses, because the effects depend on "hard-to-predict business decisions".

OECD (May 2026): judgement and responsibility

The OECD's AI exposure measure, which maps its AI capability indicators to occupations, concludes that current AI capabilities are closest to occupations involving routine information processing, administrative work and codifiable tasks. The occupations furthest away are those requiring contextual judgement, interpersonal understanding, complex decision-making and responsibility. It adds that some jobs are more exposed to robotics than to chatbots: physical does not mean untouchable.

Bank of Canada (August 2026): slower hiring, not mass layoffs

A Bank of Canada staff article, Early signs of AI-driven adjustments in Canada's labour market, found that AI "has not yet led to broad changes in the overall structure of the labour market". Exposure is highest for data entry clerks, receptionists and accounting-related roles, and lowest for nurses, judges, teachers, electricians and carpenters. The main pressure point is hiring: the estimated gap in job-finding rates between fully exposed and unexposed workers widened from 2.2 percentage points (2015 to 2019 average) to 13.9 points in 2025, while separation rates barely moved. The authors call these "only signals", stress that COVID-19, immigration and trade shifts also played a role, and say the results do not show that AI alone caused the differences.

US Bureau of Labor Statistics (August 2026): where jobs will grow

The BLS Employment Projections 2025–2035, released on 27 August 2026, expect healthcare support and healthcare practitioner occupations to grow fastest, accounting for almost one-third of all new jobs. Solar photovoltaic installers (+36.5%) and wind turbine service technicians (+29.5%) are among the fastest-growing occupations, though from a small base. Office and administrative support is projected to shrink fastest (-4.0%), with the BLS citing automation tools, including AI-powered ones.

Why some careers can't be automated easily

Five reasons keep coming back. The more of them a job combines, the more the human stays at the centre.

Two nuances. These jobs will still change: electricians will use AI for quotes, nurses for notes, mechanics for fault finding. And robotics is progressing, so the protection comes from unpredictability, not from physical work as such. A repetitive task on a production line is far more exposed than a repair in someone's kitchen.

The list: 60+ AI-proof jobs and trades AI can't replace

Each job below links to a G-Learn job trial: a 13-page guide and a short course of 5-minute lessons that show you the job from the inside, so you can test your interest before committing to a long training programme.

Construction and energy

Every building is a one-off: old wiring behind a 1950s wall, a roof pitch nobody drew, a client who changes their mind on site. These trades combine strength, fine motor skill and judgement in places robots still find very hard to work in.

• Building electrician: you sign off installations that must be safe, and you work inside real walls, not tidy diagrams.
• Plumber: every leak is in a different place, behind a different cabinet, in a home with people living in it.
• HVAC technician: diagnosing a heat pump means listening, measuring and crawling into lofts and plant rooms.
• Solar PV installer: working at height on varied roofs, with electrical safety rules a person must answer for.
• Wind turbine technician: climbing, inspecting and repairing inside a nacelle, often in poor weather; software flags faults, people fix them.
• Mason and builder: reading the ground, the old structure and the weather, then laying it true by hand.
• Roofer and carpenter: heavy, precise work at height on structures that are never quite square.
• Joiner: wood moves and every fitted piece has to adapt to the room it goes into.
• Tiler: uneven floors, awkward cuts and a finish that the client judges with their own eyes.
• Painter and decorator: surface preparation, furniture, occupied homes and a lot of small decisions on the spot.
• Drywall installer and plasterer: a smooth wall is a feel in the hand and wrist, built over years of practice.
• Glazier: handling large, fragile panes safely in places that were not designed for machines.
• Facade insulation installer: retrofit work on existing buildings, each with its own defects and access problems.
• Chimney sweep and flue fitter: inspection, safety checks and fitting in tight, dirty, irregular spaces.
• Locksmith and metalworker: call-outs to locked-out people, plus made-to-measure metalwork for real openings.
• Welder: quality welds in awkward positions on site still depend on a trained hand and eye; robots dominate repetitive factory welds, not field repairs.
• Rope access technician: inspection and repair where neither scaffolding nor machines can easily reach.
• Plant and crane operator: a busy site full of people, with accountability for every lift.
• Fibre optic technician: splicing and connecting in street cabinets, ducts and customers' homes.
• Water and sewer network technician: old, buried and poorly mapped networks that people have to find, open and fix.
• Lift engineer: safety-critical maintenance in shafts and machine rooms, with strict responsibility.
• Pool technician: water chemistry, equipment repair and regular visits to clients' properties.
• Property diagnostics surveyor: on-site inspection and a signed report someone relies on to buy or sell.
• Stonemason: reading the stone, cutting and laying it to the millimetre.

Mechanics and maintenance

Diagnostic software is now standard in every workshop, which is exactly why skilled technicians are more useful, not less: the tool points to a fault, a person confirms it, takes the vehicle apart and puts it back together safely.

• Electric vehicle technician: high-voltage safety and new systems that need trained, certified hands.
• HGV mechanic: heavy, dirty, varied repairs that keep goods moving.
• Motorcycle mechanic: one of the occupations Anthropic's data found with no measurable AI usage at all.
• Vehicle body repairer and painter: every dent and colour match is unique and judged by eye.
• Aircraft maintenance mechanic: strict regulation and a personal signature on airworthiness work.
• Agricultural machinery technician: repairs in the field, in season, on increasingly high-tech machines.
• Industrial maintenance technician: the more automated a factory becomes, the more it needs people who can fix the automation.

Health and care

Healthcare is where AI will probably help the most behind the scenes (notes, imaging, triage) and where the human part of the job stays most visible: touching, reassuring, deciding and being accountable.

• Registered nurse: clinical judgement at the bedside and a relationship of trust with the patient.
• Midwife: physical care, quick decisions and emotional support at one of life's most intense moments.
• Physiotherapist: assessment and treatment through movement and touch, adapted session by session.
• Occupational therapist: helping a person live with a condition in their own home, with their own routines.
• Speech and language therapist: tools can support practice, but therapy rests on observation and a relationship over time.
• Podiatrist: hands-on care, often for older or diabetic patients who need regular follow-up.
• Dispensing optician: measuring, fitting and adjusting frames on a real face, plus advice people trust.
• Care assistant: washing, feeding, moving and comforting people: tasks that are physical and deeply personal.
• Home care assistant: daily support in people's homes, which is very hard to automate and in growing demand as populations age.
• Ambulance crew member: emergencies in unpredictable places, under pressure, with a human life in your hands.
• Dental technician: shading, layering and fitting a tooth takes a trained eye and hands.
• Osteopath: palpating, adjusting each move and reassuring remain hands-on work.
• Psychomotor therapist: care that works through the body, presence and the relationship.
• Adult dietitian & nutritionist: the app calculates, the dietitian listens, understands and supports.
• Paediatric dietitian & nutritionist: AI can work out a menu, not a family's trust.

Social work, early years and life events

Some jobs exist precisely because a person needs another person. Parents, families and vulnerable adults expect someone who is present, accountable and able to read a situation that is not written down anywhere.

• Early years educator: young children learn through play, attention and care that only people can give.
• Registered childminder: parents entrust their child to a named, registered adult, not to a system.
• Specialist social educator: supporting people with disabilities or in difficulty requires patience, presence and trust.
• Funeral director: organising a farewell with grieving families calls for tact, care and on-the-ground logistics.
• Driving instructor: anticipating, reassuring and using the dual controls stays human work.
• Wedding & event planner: the big day is run on site, with calm nerves and trust.
• Babysitter & nanny: watching over, comforting and handling the unexpected stays human work.
• Doula: a human presence alongside care, from pregnancy to coming home.

Craft and food

Taste, texture, freshness and the pleasure of a shop counter are hard to put into an algorithm. AI may help with stock, pricing and marketing; the making and the conversation stay human.

• Baker and pastry chef: dough reacts to the temperature and humidity of the day; skilled hands adapt.
• Chocolatier: tempering, moulding and creating by hand for customers who come for craft.
• Butcher: cutting to order and advising customers across the counter.
• Fishmonger: judging freshness and preparing fish to order, every morning.
• Commis chef: cooks appear in Anthropic's zero-usage group: kitchens run on speed, teamwork and taste.
• Barista: bartenders are in that same zero-usage group; a good café runs on skill and hospitality.
• Wine merchant and sommelier: tasting, storytelling and reading what a customer will enjoy.
• Florist: composing by hand for weddings, funerals and everyday moments.
• Upholsterer: restoring one-off furniture, piece by piece.
• Cobbler and shoe repairer: repair work that is cheaper and greener than replacing, and very manual.
• Tailor and alterations specialist: fitting clothes to a real body is a craft of measuring, pinning and adjusting.
• Watchmaker & repairer: a watch is stripped, cleaned and regulated by hand.
• Cheesemonger: smelling, cutting and advising remain deeply human skills.
• Crochet bag maker: the hand, the yarn and the customer's trust stay human.
• Restaurant owner: welcoming, tasting and holding a rush together stay human skills.

Beauty and wellbeing

Clients choose these services for the hands, the eye and the conversation. Apps can book the appointment; they cannot do the cut.

• Hairdresser and barber: precise handwork on a moving head, plus a relationship that keeps clients coming back.
• Beauty therapist: treatments that rely on touch, hygiene and client trust.
• Tattoo artist: permanent work on skin, designed with the client and done by hand.
• Personal trainer: an app can write a plan; a trainer corrects posture, motivates and adapts on the day.
• Nail technician: a nail is prepared, built and decorated by hand.
• Reflexologist: touch, listening and a clear frame remain human work.
• Psychic medium: AI draws cards; a person listens, frames and protects.

Nature and animals

Outdoor and animal work is about living things in changing conditions. The Microsoft study placed grounds maintenance and forestry work among the occupational groups with the lowest AI applicability.

• Landscape gardener: design meets heavy outdoor work on sites that are all different.
• Arborist and tree surgeon: climbing, cutting and assessing risk on living trees.
• Organic market gardener: growing by observation, season after season, and selling face to face.
• Dog and cat groomer: animals do not sit still for robots; calm hands and patience matter.
• Veterinary nurse assistant: handling, caring for and reassuring animals and their owners.
• Farrier: a horse is settled, a foot trimmed, a shoe forged by hand.
• Beekeeper: hands in the apiary and an eye no sensor can replace.
• Forestry worker & lumberjack: reading each tree and deciding how it falls, out in the forest.
• Hydroponic grower: automation runs the greenhouse, eyes and hands bring in the harvest.
• Dog sitter & pet sitter: a lead, a set of keys and trust cannot be outsourced.

Jobs that work with AI

Not every resilient career is manual. A second group grows because of AI: people who build it, secure it, check it, or use it to serve clients better. The honest caveat: these roles change fast, so you need to enjoy learning continuously.

• AI engineer and consultant: companies need people who can choose, deploy and supervise AI tools responsibly.
• Cybersecurity analyst: more AI means more attack surface; humans investigate, decide and answer for incidents.
• Renewable energy engineer: designing projects that combine software, regulation and real-world sites.
• Risk manager: someone has to own the decision when models disagree or regulators ask questions.
• Real estate agent: AI drafts listings; people visit homes, negotiate and handle emotional decisions.
• B2B sales representative: routine sales tasks are being automated, so the value moves to relationships and complex deals.
• Photographer: image generation competes with stock photos, but not with being at the wedding or the event.
• Freelance entrepreneur: using AI as leverage to run a one-person business with real clients.

Browse all job trials Find my match with the quiz

The most exposed jobs: an honest look

A page about jobs AI won't replace would be misleading without the other side. The same studies agree on where AI is already doing a large share of the work:

• Data entry and administrative support. Top of the exposure lists for Anthropic and the Bank of Canada, and the fastest-shrinking group in the BLS projections.
• Customer service. High in Anthropic's observed usage and cited by the Bank of Canada as a typical high-exposure job for young workers.
• Programming. The most covered occupation in Anthropic's data, at about 75% of tasks.
• Writing, translation, media and communication. Among the highest AI applicability scores in the Microsoft study.
• Accounting support and bookkeeping. Among the highest-exposure roles in Canada.

Exposed does not mean doomed: exposure measures which tasks AI can touch, not whether employers will cut jobs. People already working as an accounting assistant, software developer, data analyst, graphic designer, UX/UI designer or community manager can stay relevant by moving towards judgement, client contact and supervising AI output. But if you are choosing a new direction from scratch, it is fair to weigh the signal that entry-level hiring in these areas has slowed.

How to choose a future-proof job from this list

A resilient job you dislike is not a good plan. Use these filters:

1. Your body and your energy. Roofing, care work and HGV repair are physically demanding. Be honest about heights, night shifts and lifting.
2. People or things? Some love the quiet focus of a workshop; others need daily contact. Health, social and beauty jobs are relationship-heavy; mechanics and many trades are less so.
3. Training time and cost. A care assistant role can be entered relatively quickly, while nursing, physiotherapy or midwifery require a degree and registration. Check the official route in your country.
4. Local demand. Look at job boards near you for the trade you are considering. Energy, building retrofit and care often have strong demand, but this varies by region.
5. Test before you commit. Read about the real daily tasks, talk to people in the job and, if possible, spend a day or two on site. Our guide on how to try a job before you switch explains the options.

If you are changing careers later in life, our guide to career change at 30, 40 or 50 covers money, timing and retraining. For the full step-by-step method, start with the complete career change guide.

Take the free career quiz

Frequently asked questions

Keep reading

Sources

• Anthropic, Labor market impacts of AI: A new measure and early evidence (March 2026)
• Tomlinson et al., Microsoft, Working with AI: Measuring the Applicability of Generative AI to Occupations (arXiv 2507.07935)
• OECD, The OECD AI exposure measure: Mapping the OECD AI Capability Indicators to occupations (May 2026)
• Bank of Canada, Early signs of AI-driven adjustments in Canada's labour market (August 2026)
• US Bureau of Labor Statistics, Employment Projections 2025-2035 Summary (27 August 2026)
• US Bureau of Labor Statistics, Fastest growing occupations