Free test
Take the testCareersGuidesArticlesMethodAbout

Method

How the AI-risk score works

Your score is an estimate of how much of your working week overlaps with what current AI systems can already do. Here is exactly how it is built, and what it can't tell you.

Last updated October 20, 2026

We score tasks, not job titles

Researchers who study AI and work, including teams at OpenAI, Anthropic, Microsoft and the U.S. Bureau of Labor Statistics, measure exposure task by task, because a job title tells you little about what someone actually does all week. Two marketing managers can have very different weeks, and very different risks.

So the test asks how much of a typical week you spend in each of 8 task areas, from "almost never" to "most of my time". Each area has an exposure rating: our estimate of the share of that kind of work current AI systems can already do or sharply speed up.

  1. Admin & data entryPaperwork, bookings, updating records90%High

    Rule-based, repeatable work on structured information is the easiest kind to automate, and it is being automated first.

  2. Writing & summarizingEmails, reports, posts, documentation85%High

    Drafting and summarizing text is the thing language models do best. A first draft that took an hour now takes a minute.

  3. Routine questionsChat, email, phone, help desk80%High

    Most routine questions have a known answer, and chat assistants already handle a large share of them without a human.

  4. Numbers & analysisReports, analysis, forecasts, looking things up70%High

    AI can now clean data, build formulas and produce a decent first analysis. Routine reporting is highly exposed; deciding what the numbers mean for the business is much less so.

  5. Digital productionCode, designs, images, video, slides65%High

    AI generates usable code, images and layouts in seconds. Production speed is no longer scarce. Taste, direction and knowing what to build still are.

  6. Leading & decidingManaging a team, signing things off, owning outcomes30%Low

    Someone has to be accountable. AI can advise, but it cannot be responsible, so decision-making roles are among the last to be automated.

  7. Face-to-face people workCaring, teaching, selling, negotiating, coaching25%Low

    Trust, persuasion and care depend on a person being present. AI can prepare and support this work, but people still want a human across the table.

  8. Hands-on physical workRepairing, building, cooking, operating equipment, treating people10%Low

    Physical work in messy, changing environments is extremely hard to automate. Robots are still far from doing it cheaply and reliably.

Share of each task area that current AI systems can already do or speed up sharply. These are the ratings the PathProof score uses. How we rate tasks

How your score is calculated

  1. Weighted task mix. We multiply the time you spend in each task area by its exposure rating, and divide by your total time. If half your week is writing (85%) and half is face-to-face work (25%), your weighted mix is 55. This mix makes up 90% of the score, so 55 contributes about 50 points.
  2. Routine adjustment. Work that is almost identical every time is easier to automate than work that is rarely the same twice. Highly repetitive work adds up to 12 points; work that is rarely the same twice takes 6 off.
  3. Presence adjustment. Work that has to happen in a particular place, with particular people or equipment, is harder to hand over to software. Work you could do entirely from a laptop adds 8 points; work you must do in person takes 10 off.
  4. Bounds. The final score runs from 4 to 96. We never show 0 or 100, because no estimate of the future deserves that confidence.

Scores fall into four bands: low (up to 29), moderate (30 to 49), high (50 to 69) and very high (70 and above).

How careers are matched

Your career matches use Holland's six interest types (practical, investigative, creative, social, enterprising and structured), the model behind most professional career guidance. Twelve quick reactions build your interest pattern, which we compare with the profile of each of the 50 careers we track.

We then adjust each match for what you told us you want: the pay, security or flexibility you care about most, the setting you prefer, how long you are willing to train, and whether the career builds on your current field. Careers with higher AI resilience get a small extra weight.

Each career's own AI exposure (0 to 100) is rated the same task-based way, from the typical task mix of the role. You can see all of them in the AI risk index.

What the ratings are based on

The exposure ratings are our own judgment, informed by published research on which tasks current AI systems perform well, including:

Pay bands are approximate U.S. figures and training times are typical routes for a newcomer. Both vary by region and experience.

What the score can't tell you

An estimate, not a prediction

Exposure is not the same as job loss. A high score means more of your current tasks overlap with what AI can do. Whether that leads to fewer jobs, changed jobs or more productive jobs depends on your employer, your industry and choices you make. The BLS makes the same point about its own categories.

  • AI is moving fast. Ratings reflect what current systems do well, and will rise for some tasks over time. We review them when major new research or official data comes out.
  • Your answers drive everything. The score is only as good as your estimate of your own week.
  • It is not advice. PathProof is an educational tool, not professional career, financial or psychological advice. Check local requirements and pay before committing to training.

Your answers stay on your device

The test runs in your browser. Your answers and your result are stored on your own device and are not sent to us. You don't need an account or an email address to see your score. See our privacy page for the details.