HOW TO READ THIS GUIDE

Clear sources. Honest uncertainty.

Useful career decisions start with understanding what a number can—and cannot—tell you.

What the ranking measures

We reproduce the ten occupations in Figure 3 of Anthropic’s March 2026 study, in descending order. Observed exposure combines task capability with work-related Claude use and weights automation more heavily.

What it does not measure

It is not a probability of dismissal, a forecast of vacancies or a measure of the entire AI market. US occupational categories and Claude usage cannot be treated as a representative survey of Brazilian workers.

How to read salaries

We use BLS OEWS data for May 2025, covering US wage and salary workers. Mean is the arithmetic average; median divides the distribution in half. P10 and P90 are percentiles, not starting pay and a career ceiling. Figures are annual USD before taxes and exclude self-employed workers. We do not convert US pay into Brazilian labor-market estimates.

Employment and opportunities

BLS projects employment change from 2025 to 2035. Annual openings include growth and replacement of workers changing occupations or leaving the workforce. They are projections, not current job ads. An occupation can shrink while still having substantial replacement openings. External vacancy links are searches only: we have not verified the returned listings.

Real occupations, stable references

Every profile maps to a SOC and O*NET code. Tasks draw on O*NET 31.0; deliverables, tools, exercises and AI interpretation are editorial adaptations. Occupations outside the ten from the study have no exposure measurement in this selection. Missing measurement does not mean no impact.

How we suggest alternatives

Each suggestion links two catalogued occupations and compares tasks, transferable skills, typical education and experience. We explain what transfers, what needs learning and the barriers involved. These are editorial hypotheses to investigate, without compatibility scores or employment promises. An alternative is not necessarily less exposed to AI, better paid or easier to enter.

Language and labor market are separate

Portuguese and English present the same United States dataset. Language changes explanations and formatting, not the country behind the numbers. Brazil will be published only after occupation and pay data are validated against national sources. Exchange-rate conversions cannot fill this gap.

Several futures are possible

Anthropic’s scenario explorer models different assumptions about adoption and capability. It is a way to think about uncertainty, not a prediction for a particular worker.

Primary sources

This site uses information from O*NET® 31.0, from the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA), under CC BY 4.0. Content has been translated, summarized and supplemented. USDOL/ETA has not approved, endorsed or tested these modifications. CC BY 4.0 ↗

Editorial review: September 10, 2026. Ranking study: March 5, 2026; usage periods: August and November 2025.

An independent educational project. Not affiliated with or endorsed by Anthropic.