Jobs That AI Can't Replace: What the UK Data Actually Shows
The 30 UK occupations least exposed to AI
| Rank of 365 | Occupation group (SOC 2010) | Skill level | Rank for LLMs | ILO flag |
|---|---|---|---|---|
| 365 | Sports players (3441) | 3 | 348 | |
| 364 | Roofers, roof tilers and slaters (5313) | 3 | 364 | |
| 363 | Elementary construction occupations (9120) | 1 | 361 | |
| 362 | Plasterers (5321) | 3 | 360 | |
| 361 | Steel erectors (5311) | 3 | 363 | |
| 360 | Vehicle valeters and cleaners (9236) | 1 | 362 | |
| 359 | Hospital porters (9271) | 1 | 306 | |
| 358 | Cleaners and domestics (9233) | 1 | 305 | |
| 357 | Floorers and wall tilers (5322) | 3 | 357 | |
| 356 | Metal plate workers, and riveters (5214) | 3 | 359 | |
| 355 | Launderers, dry cleaners and pressers (9234) | 1 | 355 | |
| 354 | Window cleaners (9231) | 1 | 309 | |
| 353 | Painters and decorators (5323) | 3 | 337 | |
| 352 | Fork-lift truck drivers (8222) | 2 | 365 | High augmentation |
| 351 | Packers, bottlers, canners and fillers (9134) | 1 | 338 | |
| 350 | Gardeners and landscape gardeners (5113) | 3 | 340 | |
| 349 | Bricklayers and masons (5312) | 3 | 346 | |
| 348 | Road construction operatives (8142) | 2 | 353 | |
| 347 | Elementary process plant occupations n.e.c. (9139) | 1 | 350 | |
| 346 | Tyre, exhaust and windscreen fitters (8135) | 2 | 342 | |
| 345 | Industrial cleaning process occupations (9132) | 1 | 351 | |
| 344 | Street cleaners (9232) | 1 | 347 | |
| 343 | Vehicle paint technicians (5234) | 3 | 358 | |
| 342 | Butchers (5431) | 3 | 326 | |
| 341 | Farm workers (9111) | 1 | 339 | |
| 340 | Dancers and choreographers (3414) | 3 | 263 | |
| 339 | Process operatives n.e.c. (8119) | 2 | 343 | |
| 338 | Textile process operatives (8113) | 2 | 345 | |
| 337 | Scaffolders, stagers and riggers (8141) | 2 | 341 | |
| 336 | Large goods vehicle drivers (8211) | 2 | 354 |
Rank 1 = most exposed. Source: Department for Education (Nov 2023): The impact of AI on UK jobs and training, Annex 1, AI Occupational Exposure method (Felten, Raj and Seamans, 2023). Contains public sector information licensed under the Open Government Licence v3.0. Exposure is not a prediction of job loss. Check your own role's tasks.
Why some jobs resist AI: the three core reasons
AI is very good at pattern recognition, text generation, and processing structured data. It struggles with three things that many jobs depend on.
- Physical dexterity in unpredictable spaces. Fitting a pipe under a Victorian terrace, tiling an uneven wall, or replacing a roof in a British November involves constant micro-adjustments that current robotics cannot replicate cheaply or reliably.
- High-stakes human trust. A nurse assessing whether a patient is deteriorating, a paramedic calming a frightened child, or a social worker reading a family situation draws on emotional intelligence and contextual judgement that AI cannot safely replace [4].
- Professional and legal accountability. Surgeons, barristers, and registered engineers make decisions where someone must hold a licence and be legally responsible. AI can assist, but it cannot stand in a court or carry a professional registration.
The Department for Education's 2023 report on AI and UK jobs uses an AI Occupational Exposure (AIOE) score to rank all 365 UK occupation groups [2]. A low score means the abilities the job requires overlap little with what current AI applications can do. It does not mean the job is completely untouched — it means the overlap is small compared with other roles [3].
The UK government's list of jobs least exposed to AI
The DfE report's Table 3 ranks the 20 UK occupations with the lowest AI exposure scores [3]. Here they are, in official rank order, with a brief note on why each scores low.
| Rank (1 = least exposed) | Occupation | Why exposure is low |
|---|---|---|
| 1 | Sports players | Physical performance, split-second decisions, live crowd interaction |
| 2 | Roofers, roof tilers and slaters | Outdoor physical work, variable conditions, safety-critical |
| 3 | Elementary construction occupations | Manual labour in unpredictable environments |
| 4 | Plasterers | Fine motor skill, surface judgement, physical presence |
| 5 | Steel erectors | Height, physical strength, real-time problem-solving |
| 6 | Vehicle valeters and cleaners | Physical, variable, low structured-data content |
| 7 | Hospital porters | Physical movement, human interaction, time-sensitive |
| 8 | Cleaners and domestics | Physical, variable environments, human presence required |
| 9 | Floorers and wall tilers | Precision manual skill, bespoke fitting |
| 10 | Metal plate workers and riveters | Specialist physical fabrication |
| 11 | Launderers, dry cleaners and pressers | Physical handling, garment judgement |
| 12 | Window cleaners | Physical, outdoor, access-dependent |
| 13 | Painters and decorators | Surface preparation, colour matching, physical skill |
| 14 | Fork-lift truck drivers | Physical operation, spatial awareness |
| 15 | Packers, bottlers, canners and fillers | Repetitive but physical and variable |
| 16 | Gardeners and landscape gardeners | Outdoor, seasonal, creative and physical |
| 17 | Bricklayers and masons | Precision manual craft, structural judgement |
| 18 | Road construction operatives | Physical, outdoor, safety-critical |
| 19 | Elementary process plant occupations n.e.c. | Physical operation, variable conditions |
| 20 | Tyre, exhaust and windscreen fitters | Physical, diagnostic, customer-facing |
The pattern is consistent: almost every role involves being somewhere physical, doing something with your hands, in conditions that change. That is exactly what AI finds hard [3].
These rankings come from the DfE's AIOE methodology, which measures how much the abilities a job requires overlap with current AI capabilities. A low score does not mean zero exposure — it means the overlap is smaller than for other roles [3].
Jobs safe from AI: care, health and emergency services
The DfE data focuses on exposure scores across all occupations. The Oxford University AI in Education project highlights a second cluster of roles that are safer for different reasons: they require physical presence, human trust, and ethical accountability [4].
- Nurses and healthcare assistants. Patient assessment, emotional support, and physical care all require a human in the room. AI can flag a deteriorating patient from monitor data, but it cannot hold a hand or make a clinical call.
- Paramedics and emergency responders. Every call is different. Improvisation in a stranger's home, under time pressure, with incomplete information, is not a task AI can reliably perform [4].
- Physiotherapists and occupational therapists. Hands-on treatment, real-time adjustment based on patient feedback, and therapeutic relationship all resist automation [4].
- Electricians and plumbers. Oxford's AI in Education team specifically names these trades [4]. Fault-finding in an old house, working in confined spaces, and meeting Part P or Gas Safe registration requirements all need a licensed human.
- Cybersecurity specialists. Attackers adapt constantly. Defending against novel threats requires creative, adversarial thinking that AI tools support but cannot replace [4].
- Teachers and early years practitioners. Behaviour management, pastoral care, and the relationship between teacher and pupil are deeply human. AI can personalise content; it cannot replace the adult in the room.
- Social workers. Safeguarding decisions carry legal weight and moral complexity. A human professional must make and own those calls.
The DfE report also identifies the least exposed sectors overall as accommodation and food services, motor trades, agriculture, forestry and fishing, transport and storage, and construction [3]. If you work in one of these sectors, your exposure is likely lower than average — though individual roles within any sector still vary significantly.
What makes a job that AI can't replace: a practical checklist
Use this checklist to assess your own role or a role you are considering. The more boxes you tick, the lower your likely exposure.
- ☐ The work happens in a physical location that changes (a patient's home, a building site, a vehicle)
- ☐ You regularly deal with situations you have never encountered in exactly that form before
- ☐ Your decisions carry professional, legal or ethical accountability
- ☐ Clients or patients specifically need to trust *you*, not just a service
- ☐ The job requires a licence, registration or regulated qualification (Gas Safe, NMC, SRA, HCPC)
- ☐ Physical dexterity or strength is central to the role
- ☐ You read emotional cues and adapt your approach in real time
- ☐ The output cannot be verified by a machine before it reaches a person
If you are unsure where your current role sits, the AI Threat Score gives you a personalised exposure estimate in about 60 seconds — no sign-up needed. It is an estimate, not a prediction.
Jobs AI cannot replace: what this means for your career decisions today
Knowing a role is low-exposure is useful, but it is not a reason to stop developing. Here is what the evidence suggests you should actually do.
- Audit your current skills honestly. Which parts of your job involve the checklist factors above? Which parts are more routine and data-based? A Skills Audit can help you map this clearly.
- Build the human skills that AI cannot replicate. Emotional intelligence, physical craft, ethical reasoning, and licensed expertise are the moat. Invest in CPD, apprenticeships, or professional registration if you have not already.
- Add AI literacy on top, not instead. Workers who do best in low-exposure roles often use AI tools to handle administrative overhead — documentation, scheduling, reporting — while keeping the human core of their work. A nurse who uses AI to reduce paperwork has more time for patients.
- Make sure your CV reflects the human value you add. If your CV reads like a list of tasks, it undersells the judgement, trust, and physical skill that make you hard to replace. Review each bullet point and ask whether it shows a decision you made, not just a duty you performed. CV Studio can help you reframe your experience clearly.
- Stay current on your sector's exposure. AI capabilities are changing. A role that is low-exposure today may look different in five years. Set a reminder to reassess annually.
- If you are considering a career change, use the data. The DfE's sector rankings [3] are a reasonable starting point. Construction, health and social care, and skilled trades all sit at the lower end of exposure. Compare that with your interests and the job market in your area.
The DfE report notes that regional differences in AI exposure are much smaller than differences between occupations and industries [3]. Your job title matters far more than where you live.
A note on the jobs most exposed to AI
Understanding which jobs AI can't replace is easier when you see the contrast. The DfE's Table 2 lists the 20 most-exposed UK occupations, dominated by management consultants, financial managers, chartered accountants, actuaries, solicitors, and HR administrators [3]. These roles involve processing structured information, drafting documents, and analysing data — exactly what large language models handle well.
These roles are not necessarily disappearing, but they are changing fastest. If your role sits closer to that end of the spectrum, see Jobs That Will Be Replaced by AI: What the UK Data Actually Says for a practical guide to that situation.
For a broader overview of how to position yourself across the full landscape, AI-Proof Jobs: What They Are and How to Find Yours covers the wider picture, including how to identify and strengthen the parts of your current role that are hardest to automate.
Frequently asked questions
What five jobs will AI not replace?
Based on UK government data, the five least-exposed occupations are sports players, roofers, elementary construction workers, plasterers, and steel erectors [3]. Beyond trades, nurses, paramedics, electricians, and physiotherapists are also consistently cited as low-exposure because they combine physical presence, licensed accountability, and human trust in ways AI cannot replicate [4].
Are skilled trades genuinely safe from AI?
Safer than most, yes. The DfE's 2023 report places construction trades, vehicle maintenance, and cleaning occupations at the bottom of its AI exposure ranking [3]. These roles involve physical work in variable, unpredictable environments — exactly where current AI and robotics struggle most. That said, administrative parts of any trade role, such as quoting or invoicing, carry more exposure.
Which sectors are least exposed to AI in the UK?
The DfE identified accommodation and food services, motor trades, agriculture, forestry and fishing, transport and storage, and construction as the least exposed sectors [3]. Finance and insurance is the most exposed, followed by information and communication and professional, scientific and technical services. Individual roles within any sector still vary significantly.
Does having a degree make my job more exposed to AI?
It can. The DfE report found that higher-skill occupations — those typically requiring degree-level qualifications — tend to appear more frequently in the high-exposure rankings, because degree-level work often involves structured information processing, writing, and analysis that large language models handle well [3]. Physical and care-based roles, which often do not require a degree, tend to score lower.
How do I find out my personal AI exposure?
The AI Threat Score gives you a personalised exposure estimate based on your job title and sector. It takes about 60 seconds, needs no sign-up, and is free. It is an estimate, not a prediction — a starting point for thinking about where your role sits relative to others.
Should I retrain into a low-exposure job if mine is high-exposure?
Not necessarily. The more practical first step is to identify which parts of your current role are low-exposure — the human judgement, relationship, or physical elements — and build those up. Retraining is worth considering if your sector is shrinking, but adding AI literacy and strengthening the human core of your existing role is often faster and more effective.
Related guides
- AI-Proof Jobs: What They Are and How to Find Yours
Which jobs are least at risk from AI in the UK? Understand what makes a role AI-proof, which are most exposed, and how to check your own position.
- Jobs That Will Be Replaced by AI: What the UK Data Actually Says
Which jobs that will be replaced by AI are most at risk in the UK? See the official data, what's changing by 2030, and steps to protect your career today.