Which jobs go underwater, and when
Each bubble is a profession, sized by its workers. Across: the year AI can first do the job. Up: years of friction before it actually does.As the waterline rises, jobs go under. The waterline is diagonal reflecting that jobs higher up in the chart have more friction and AI replaces them years after it is capable of doing so. Solid line: today. Dashed: 2030, 2033 and 2036. ▶ watch the water rise. How to read it
Swipe sideways for later years
Professions
Employment from ILO, Eurostat and BLS. Readiness, friction and headroom are judgment estimates; hover over (or tap) a row for the reasoning. Sorted by the year AI takes over under the current dials. Download all data (Excel): workers by region, the judgments, the years they imply across the five scenarios (earliest, steady, latest), the reasoning, and which sliders move each row. Edit the amber columns and upload your assumptions to redraw the charts with them; they stay in your browser.
Jobs automated since today, by kind of work
Job-equivalents automated beyond today's level, net of reabsorption. Hover for values.
Share of each occupation group automated
Major SOC occupation groups, split by kind of work.
Most affected occupations
Ranked by job-equivalents automated. Mix shows each job's split of work time across thinking, driving, structured and unstructured physical tasks.
Occupation landscape
Each bubble is an occupation, sized by employment. Across: share of work time in tasks today's LLMs are exposed to. Up: share of work time that is physical. Color: share of the job automated by the selected year. Press play to watch it fill in.
Workforce map
Every US wage-and-salary job, grouped by occupation. Area is employment; color is the share automated by the selected year.
When the wave reaches each group
Share of each group's jobs automated by year, sorted by the selected year. Cognitive work moves first, and physical work follows as robotics matures.