Will AI Take My Job? Score Your Risk with 2026 Data

300 million full-time jobs worldwide are exposed to automation from today's AI, according to a 2023 Goldman Sachs estimate. That figure gets quoted to frighten people, and it shouldn't, at least not the way it usually is, because "exposed" is not the same as "gone." The same research says most of those jobs get changed, not deleted.

So the useful question isn't the panicky "will a robot take my job." It's narrower and more answerable: how much of what you actually do each day can AI already handle, and does that add up to your role being automated away, or just reshaped around you? This guide gives you a way to score that for your own job, using 2026 research instead of the 2013 numbers most online "job risk" checkers still quietly run on.

A note on where the numbers come from. Everything here is drawn from named public research: the World Economic Forum, McKinsey, Goldman Sachs, Pew Research, the OECD, and the ILO, with dates attached, not from any hands-on prediction of your workplace. Nobody can tell you for certain what your employer will do. What the research can do is show you where the pressure is real and where it's overblown.

Will AI take your job?

Whether AI takes your job depends far more on your daily tasks than on your job title. AI models are good at a specific set of activities: writing and editing text, summarizing documents, drafting and debugging code, processing data, and answering routine questions. They're weak at others: physical work, hands-on care, original judgment, and anything where a human has to be accountable for the outcome. A job is just a bundle of tasks. The more of your bundle sits in that first list, the more exposed you are.

This is why two people with the same title can face very different risk, and why "exposure" almost never means your job vanishes overnight. The OECD's 2023 Employment Outlook made the point plainly: even high-exposure occupations are unlikely to be fully replaced, because most jobs mix automatable and non-automatable work. Exposure tells you how much pressure is on the role. It doesn't tell you the role is finished.

Score your own job: the 5-factor AI risk test

The AI risk test scores your job on five task factors, each from 0 (insulated) to 3 (highly exposed), for a total out of 15. Rate your own work honestly on each row, add up the points, and read the band underneath.

Add your five scores, then find your band:

• 0 to 4, lower exposure. AI will help with pieces of your work, but the core is hard to automate. Your risk is falling behind on the tools, not losing the role.
• 5 to 9, moderate exposure. Expect parts of your job to get automated or sped up. The work changes rather than disappears, and staying valuable means owning the parts AI can't do.
• 10 to 15, higher exposure. A large share of your tasks is already within reach of current AI. That rarely means sudden redundancy, but the role tends to shrink or get redefined, so it pays to get ahead of it.

One honest caveat built into the test: a high score is a flag, not a verdict. It usually means "AI can do a chunk of your tasks," which more often produces a smaller or changed job than a deleted one. The ILO's 2023 analysis of generative AI reached the same conclusion, that its biggest effect is augmenting tasks inside jobs rather than wiping out whole occupations.

What the test looks like on real jobs

Scoring a few familiar roles shows how the factors stack up, and why the results line up with the wider research rather than gut feeling.

These match the data. The World Economic Forum's 2025 Future of Jobs report lists data-entry and administrative clerks among the fastest-declining roles, while Goldman Sachs put legal work at 44% task exposure in 2023. Nursing and skilled trades sit at the other end, which is exactly where the test lands them.

Which jobs are most exposed to AI right now?

The jobs most exposed to AI in 2026 are clerical, administrative, and information-heavy desk roles, not the factory and manual work older forecasts feared. Goldman Sachs estimated in 2023 that generative AI could take on a large share of tasks in the most information-based occupation groups.

The pattern repeats across the other research. The World Economic Forum's 2025 report names data-entry clerks, administrative and executive secretaries, bookkeeping and payroll clerks, bank tellers, and cashiers among the fastest-declining roles this decade. Pew Research pointed to budget analysts, tax preparers, technical writers, and web developers as highly exposed. The ILO found clerical work the single most exposed category, with about a quarter of clerical tasks at high exposure and well over half at medium.

Which jobs are safest from AI?

The safest jobs from AI are hands-on, physical, and high-contact human roles, where dexterity, perception, and trust matter more than processing information. Goldman Sachs put task exposure lowest in exactly these groups.

Growth backs this up, not just low risk. The World Economic Forum projects the largest absolute job gains to 2030 in farming, delivery driving, construction, and care and nursing. Pew's low-exposure examples were roles like barbers, child-care workers, and nannies. What protects all of them are the three things machines have long struggled with, described back in 2013 by Oxford's Frey and Osborne as automation bottlenecks: perception and manipulation, creativity, and social intelligence.

Why the popular "will a robot take my job" tools get it wrong

Most job-risk checkers you'll find online run on data from 2013, before large language models existed. The best-known ones, willrobotstakemyjob.com and the BBC's "Will a robot take your job?" tool, both rest on Frey and Osborne's 2013 model, which scored 702 occupations and produced the famous headline that 47% of US employment was at high risk. That model was built and trained more than a decade ago, before ChatGPT, on assumptions from a pre-AI world.

The bigger problem is that those assumptions are now partly backwards. The 2013 models treated manual and physical work as the most automatable and desk work as relatively safe. Generative AI flipped that. It's the analytical, clerical, and writing-heavy jobs that current AI touches first, while a plumber or a nurse is comparatively insulated. So a tool built on 2013 data can rate a data-entry role "safe" and a driver "doomed," which is close to the opposite of what today's AI actually does. That's the whole reason to score your tasks against current research instead of trusting a single old percentage.

Does higher pay or a degree protect you?

A degree and a high salary don't shield you from AI, and can point the other way. Pew Research found in 2023 that the most AI-exposed US jobs are the higher-paid, more educated, analytical ones. Among workers with a bachelor's degree, 27% were in the most-exposed jobs, against just 12% of those with only a high-school diploma. The exposed jobs paid a median of about $33 an hour, compared with $20 for the least-exposed.

The logic is straightforward once you drop the old intuition. AI is good at the cognitive, information-processing work that knowledge jobs are built on, so the further your work moved away from your hands and toward a screen, the more of it current models can attempt. Education bought you into that kind of work. It doesn't buy you out of the exposure.

What AI still can't do well

Four capabilities keep coming up across every study as hard to automate, and they map directly onto the safer end of the test. Physical dexterity and working in the messy real world, the thing that keeps trades and hands-on care insulated. Genuine creativity and original judgment, where the answer isn't in any training data. Social intelligence, the trust, care, persuasion, and reading of a room that human relationships run on. And accountability, the roles where a person has to own the decision because a licence, a law, or a life depends on it.

The OECD found that high-skill occupations carry the lowest automation risk overall, which fits: those are the jobs richest in judgment and human responsibility. If your work leans on these four, AI is far more likely to become a tool you use than a replacement for you.

How to lower your AI job risk

You can shift your own exposure by moving your time toward the factors that scored low. Take on more of the judgment calls, the client relationships, the physical or complex problem-solving, and less of the routine screen work a model can already draft. The point isn't to avoid AI. It's to be the person wielding it rather than the person it replaces.

Learning to use the tools well is the most direct move, and it's how the fastest-growing roles are defined: the World Economic Forum's 2025 list is topped by AI and machine-learning specialists, data analysts, and information-security experts. You don't need to become an engineer to benefit. If you're new to it, our guide on learn to use the tools yourself is a starting point, and the AI tools worth knowing covers what's actually useful by job. If you're weighing a move, using AI to run your job search is the practical companion to this piece.

Does the answer change by country?

The task pattern holds worldwide, but the pressure lands on different jobs depending on the local economy. The global figures here from the World Economic Forum and Goldman Sachs apply broadly, because a routine task is routine everywhere.

In India, the exposure concentrates on the IT-services and business-process outsourcing sectors that employ millions, with manual software testing, first-line support, and data processing most in the firing line. The industry body NASSCOM frames this as roles being transformed rather than erased, and expects new AI-related jobs to be created alongside the disruption, a more measured read than some of the gloomier domestic forecasts. In the UK and other Western markets, you'll still see a recycled figure that around 30% of jobs could be automatable by the early 2030s, but that estimate comes from PwC's 2017 analysis, which predates generative AI, so treat any pre-2020 percentage with caution. Newer, task-based numbers are the ones to trust.

What this guide does not cover

This is an estimate to help you think, not a forecast of your future or professional career advice. The score is a rough exposure gauge built from public research, and your real situation depends on your employer, your industry, and how fast the tools get adopted where you work. For a major career or financial decision, talk to a qualified advisor rather than an article. It also doesn't cover specific company plans or salaries. If you'd rather answer the five factors and get a score automatically, our AI job risk calculator runs the same test interactively. Either way, the five factors are enough to place your own job honestly and decide what to do about it.

Sources

1. World Economic Forum: Future of Jobs Report 2025
2. Goldman Sachs: Generative AI could raise global GDP by 7% (2023)
3. Pew Research Center: Which U.S. workers are more exposed to AI on their jobs? (2023)
4. McKinsey Global Institute: Generative AI and the future of work in America (2023)
5. OECD Employment Outlook 2023: Artificial intelligence and jobs
6. ILO: Generative AI and jobs, a global analysis (2023)