Is Your Job at Risk from AI? How to Read Your Exposure

If you have typed “will AI take my job” into a search bar, here is the honest answer in two sentences. AI exposure measures how much of your job’s tasks current AI could plausibly do or speed up, not whether your role will be eliminated. Almost every job is a mix of exposed tasks and durable tasks, so a high exposure number is a signal to reposition, not a verdict on your career.

The panic behind the search is understandable, but it asks the wrong question. The useful question is not whether the whole job goes away. It is which tasks overlap with what AI can do today, and what you do about the ones that do. This piece gives you a plain definition of exposure, a do-it-yourself task audit you can run in an afternoon, the evidence on augmentation versus automation, and a way to get a personalized, source-traceable read on your own occupation.

Once you understand what exposure actually measures, the score stops being frightening and starts being useful. When you are ready for a personalized read, you can take the free AI Ready Score and see how your specific occupation breaks down.

What does AI exposure actually mean, and is it the same as being replaced?

AI exposure is the share of an occupation’s tasks that current AI capabilities could plausibly perform or accelerate. It is a measure of overlap between what AI can do and what your work involves. It is not a prediction of layoffs, and it does not say your employer will choose to act on it.

The term comes from the research, not from marketing. The OpenAI and University of Pennsylvania paper “GPTs are GPTs” (Eloundou, Manning, Mishkin and Rock, 2023, arXiv:2303.10130) introduced this task-level exposure measure and applied it explicitly without distinguishing labor-augmenting from labor-displacing effects. The authors built the idea to be neutral about the outcome. Exposure tells you that AI capability and a task overlap. It does not tell you whether that overlap helps you or replaces you.

The scale is deliberately broad. The same paper estimated that about 80% of US workers could have at least 10% of their tasks exposed to large language models, and about 19% could see at least 50% of their tasks exposed (OpenAI and University of Pennsylvania, 2023). When a measure touches four in five workers at some level, it cannot mean “replaced” for most people. Broad exposure is exactly why the number needs reading, not fearing.

This matters because many of the viral “will AI take my job” calculators still run on the Frey and Osborne automation probabilities from 2013. Those scored whole occupations and predate large language models entirely. Reading exposure at the task level, using current sources, gives a more honest picture. You can see the methodology and the sources behind every number that JobRoute uses to do exactly that.

Why does a high exposure score not mean your job disappears?

Because exposure concentrates in tasks, not whole jobs. Real-world usage shows AI landing on a fraction of an occupation’s tasks for almost everyone, not blanketing the entire role.

The strongest task-level evidence comes from the first Anthropic Economic Index (February 2025), which analyzed real, anonymized Claude usage rather than survey opinions. About 36% of occupations used Claude for at least a quarter of their tasks, but only about 4% used it for at least three quarters of their tasks. Whole-job coverage is rare. Most occupations show AI helping with a slice of the work, not all of it.

The same analysis examined how people used the tool. Roughly 57% of Claude usage augmented the human, working alongside the person, while about 43% automated the task outright (Anthropic Economic Index, February 2025). That is observed behavior, not a forecast. And the balance moves. In Anthropic’s January 2026 update, analyzing November 2025 data, the share of Claude.ai conversations classified as augmented rose to about 52% while automated fell to about 45%, after a period when automation had briefly led. The augmentation and automation mix is dynamic, which is another reason a single scary headline number misleads.

How do you audit your own role instead of trusting a generic label?

The occupation label is not enough. Two people with the same job title spend their weeks on different task mixes, so a personalized read beats a headline percentage every time. Here is a do-it-yourself audit you can run in one sitting.

Step 1: list your tasks. Write the 8 to 12 things you actually spend time on in a typical week, in plain verbs. Draft, review, schedule, advise, troubleshoot, persuade, decide, reconcile. Be specific to your week, not your job description.

Step 2: mark what AI can do today. For each task, mark whether current tools can do it well now, do part of it, or cannot do it reliably. Judge on quality, not just possibility. “It could generate something” is not the same as “it could do this at the standard my work requires.”

Step 3: sort exposed from durable. Tasks heavy in pattern, drafting, summarizing and structured retrieval tend to be exposed. Tasks heavy in judgment, accountability, physical presence, relationship trust and ambiguous trade-offs tend to be durable. Your two lists are now the map you needed.

A self-audit is a starting estimate, and it is worth being honest about that limit. For a source-traceable read across all of your occupation’s tasks, you can take the free AI Ready Score and see how the score works under the hood.

Which skills hold their value as AI capability rises?

The durable half of your audit, the tasks AI cannot reliably own, points directly to the skills worth protecting and deepening. And employers agree on which ones matter.

According to the World Economic Forum’s Future of Jobs Report 2025, analytical thinking is the most sought-after core skill, with about seven in ten employers calling it essential. It is followed by resilience, flexibility and agility, then leadership and social influence. These are not technical certifications. They are the judgment, adaptability and people-facing capabilities that map onto exactly the tasks you marked as hard for AI in step 3 of your audit.

These are the parts of work that compound in value as routine tasks get faster. For a deeper look at the durable skills that hold their value, the longer read goes occupation by occupation.

What should you do this week if your role looks highly exposed?

A high exposure read is the start of a plan, not the end of a career. Here are three concrete moves you can begin now.

Move 1: lean into your durable skills. Pick one or two durable tasks from your audit and deliberately take on more of that work. Those are the parts employers rate most highly, and they are where your value grows rather than erodes.

Move 2: adopt the tools on your exposed tasks. Learn to use AI on the augmentable tasks so you become the person who is faster and better with the tools, not the person replaced by them. Remember that the OpenAI and University of Pennsylvania study (2023) estimated about 15% of all worker tasks could be done significantly faster at the same quality with a large language model. That speed is yours to capture.

Move 3: scout an adjacent role. Identify roles that reuse your durable skills with lower exposure. The skills transfer; the task mix changes. You can find an adjacent role that builds on what you already do well.

The optimism here is grounded in scale, not sentiment. The dominant response to exposure across the economy is training, not termination. The World Economic Forum’s Future of Jobs Report 2025 projects that 59% of workers will need upskilling or reskilling by 2030, and forecasts 170 million new roles created against 92 million displaced, for a net gain of 78 million jobs.

What can an AI exposure score honestly tell you, and what can it not?

A score is only useful if you trust its boundaries, so here they are plainly. What an exposure score can tell you: which of your tasks overlap with current AI capability, how concentrated that overlap is, and which durable skills and adjacent roles to move toward.

What it cannot tell you: whether your specific employer will choose to automate, when that might happen, how your local labor market is moving, or how your individual performance changes the picture. Exposure is capability overlap. It is not a personal forecast.

The broader churn supports patience over panic. Skill instability is real but slowing, from 57% of skill sets transformed or outdated in 2020, to 44% in 2023, to a projected 39% by 2030 (World Economic Forum, Future of Jobs Report 2025). The trajectory rewards steady adaptation, not alarm.

This is also the difference worth naming plainly. The major studies, from the World Economic Forum to OpenAI to Anthropic, publish episodic reports, not a queryable per-occupation tool. JobRoute reconciles O*NET, the OpenAI and University of Pennsylvania exposure study, the Anthropic Economic Index, the WEF Future of Jobs Report 2025, and BLS into one source-traceable engine that names exposure, durable skills and adjacent roles in one place.

Get your personalized, source-traceable read

You now know that exposure is task-level, that real usage leans toward augmentation, and that your durable skills and adjacent roles are the levers within your control. The next step is to turn that understanding into a read on your own work.

Take the free AI Ready Score for a per-occupation read across one of 1,016 O*NET occupations, with every number traceable to its source. If you want to check the engine before you trust it, see the methodology and the sources behind every number and exactly how the score works.

Exposure is the start of a plan, not the end of a career.