Which jobs AI changes most

In 372 of 773 US occupations, AI does or helps with at least half of the work, weighting each task by how much it matters — work done by 75.8 million people. Among sizeable jobs it reaches furthest into accountants and auditors, computer systems analysts, and personal financial advisors. It touches least — the closest thing to AI-proof jobs — landscaping and groundskeeping workers, industrial truck and tractor operators, and dishwashers.

Tasks from O*NET Database 31.0 (2026), each scored with our published task-content method; employment and pay from BLS Occupational Employment and Wage Statistics, May 2025. Updated November 2026. Method

Occupations ranked
773
O*NET Database 31.0
AI does or helps with half the work or more
372
48% of occupations
People in those jobs
75.8 million
BLS employment, May 2025
Hours back a week, typical job
3–8
median of the estimated range

Every occupation, ranked

Sorted by how much of a job’s work AI does or helps with. Click a column to sort by hours, headcount or pay. A job opens its own AI skills section when it has a page here, and its learning path otherwise.

773 occupations · showing the first 50

#Occupation
1Accountants and AuditorsBusiness and Financial100%12 of 12 tasks6–16
2Computer Systems AnalystsComputer and Mathematical100%12 of 12 tasks6–16
3Personal Financial AdvisorsBusiness and Financial100%12 of 12 tasks6–16
4Data ScientistsComputer and Mathematical100%12 of 12 tasks6–16
5Cost EstimatorsBusiness and Financial100%12 of 12 tasks6–16
6Insurance Claims and Policy Processing ClerksOffice and Administrative Support100%12 of 12 tasks6–16
7Biological Scientists, All OtherLife, Physical and Social Science100%12 of 12 tasks6–16
8Title Examiners, Abstractors, and SearchersLegal100%12 of 12 tasks6–16
9Health Information Technologists and Medical RegistrarsHealthcare Practitioners100%12 of 12 tasks6–16
10Brokerage ClerksOffice and Administrative Support100%10 of 10 tasks6–16
11Correspondence ClerksOffice and Administrative Support100%12 of 12 tasks6–16
12Mathematical Science Occupations, All OtherComputer and Mathematical100%12 of 12 tasks6–16
13Urban and Regional PlannersLife, Physical and Social Science95%12 of 12 tasks6–15
14Atmospheric and Space ScientistsLife, Physical and Social Science95%12 of 12 tasks6–15
15Computer Science Teachers, PostsecondaryEducation95%12 of 12 tasks6–15
16Clinical and Counseling PsychologistsLife, Physical and Social Science95%12 of 12 tasks6–15
17Credit Authorizers, Checkers, and ClerksOffice and Administrative Support95%12 of 12 tasks6–15
18Executive Secretaries and Executive Administrative AssistantsOffice and Administrative Support95%12 of 12 tasks6–15
19Human Resources Assistants, Except Payroll and TimekeepingOffice and Administrative Support95%12 of 12 tasks6–15
20Interpreters and TranslatorsArts, Design and Media95%12 of 12 tasks6–15
21StatisticiansComputer and Mathematical95%12 of 12 tasks6–15
22Credit CounselorsBusiness and Financial95%12 of 12 tasks6–15
23Medical Secretaries and Administrative AssistantsOffice and Administrative Support95%12 of 12 tasks6–15
24Market Research Analysts and Marketing SpecialistsBusiness and Financial95%12 of 12 tasks6–15
25Claims Adjusters, Examiners, and InvestigatorsBusiness and Financial95%12 of 12 tasks6–15
26Court, Municipal, and License ClerksOffice and Administrative Support95%12 of 12 tasks6–15
27Medical TranscriptionistsHealthcare Support95%12 of 12 tasks6–15
28Physical Scientists, All OtherLife, Physical and Social Science95%12 of 12 tasks6–15
29Bookkeeping, Accounting, and Auditing ClerksOffice and Administrative Support95%12 of 12 tasks6–15
30Computer Occupations, All OtherComputer and Mathematical95%12 of 12 tasks6–15
31Production, Planning, and Expediting ClerksOffice and Administrative Support95%12 of 12 tasks6–15
32Statistical AssistantsOffice and Administrative Support95%12 of 12 tasks6–15
33Genetic CounselorsHealthcare Practitioners94%12 of 12 tasks6–15
34Credit AnalystsBusiness and Financial94%11 of 11 tasks6–15
35Tax PreparersBusiness and Financial94%12 of 12 tasks6–15
36Billing and Posting ClerksOffice and Administrative Support92%11 of 12 tasks6–15
37Medical Records SpecialistsHealthcare Practitioners92%11 of 12 tasks6–15
38Judicial Law ClerksLegal91%12 of 12 tasks5–15
39File ClerksOffice and Administrative Support91%11 of 12 tasks5–15
40Economics Teachers, PostsecondaryEducation90%12 of 12 tasks5–14
41Architectural and Civil DraftersArchitecture and Engineering90%12 of 12 tasks5–14
42Mathematical Science Teachers, PostsecondaryEducation90%12 of 12 tasks5–14
43Geography Teachers, PostsecondaryEducation90%12 of 12 tasks5–14
44Librarians and Media Collections SpecialistsEducation90%12 of 12 tasks5–14
45Education Teachers, PostsecondaryEducation90%12 of 12 tasks5–14
46Legal Secretaries and Administrative AssistantsOffice and Administrative Support90%12 of 12 tasks5–14
47Political Science Teachers, PostsecondaryEducation90%12 of 12 tasks5–14
48HistoriansLife, Physical and Social Science90%12 of 12 tasks5–14
49Payroll and Timekeeping ClerksOffice and Administrative Support89%12 of 12 tasks5–14
50Business Teachers, PostsecondaryEducation89%12 of 12 tasks5–14

AI share: each task’s exposure weighted by how important O*NET rates it — a task AI does counts in full, one it helps with counts about a third. Hours back a week: our estimated range for someone who uses AI well. Employment and median pay: BLS.

Where AI exposure is highest, by state

Across the states and DC, the share of people who work in jobs where AI touches at least half of the work — an AI exposure index of 0.50 or more — is highest in the District of Columbia, at 72.0%, and lowest in Mississippi, at 44.0%. Nationwide it is 51.8%, and AI could save an estimated 452 million–1.2 billion hours of work a week across the 155.4 million workers counted. States differ only through the mix of jobs they do: exposure and task share are measured per occupation, nationally.

#StateWork in jobs where AI touches at least half of the workHours AI could save a week
1District of Columbia72.0%484,370 workers2.5–6.6 million4–10 hours a worker
2Maryland57.9%1,539,630 workers8.4–22.3 million3–8 hours a worker
3Massachusetts57.6%2,039,370 workers11.2–29.8 million3–8 hours a worker
4Colorado56.2%1,567,270 workers8.7–23.1 million3–8 hours a worker
5Virginia55.2%2,208,670 workers12.4–33.0 million3–8 hours a worker
6Utah54.2%905,350 workers5.0–13.3 million3–8 hours a worker
7New Hampshire54.1%359,740 workers2.0–5.3 million3–8 hours a worker
8New York53.6%5,121,810 workers29.2–77.9 million3–8 hours a worker
9Connecticut53.6%872,590 workers4.9–13.0 million3–8 hours a worker
10Texas53.4%7,405,610 workers41–110 million3–8 hours a worker
11Florida53.3%5,159,090 workers29.0–77.3 million3–8 hours a worker
12Washington53.3%1,849,450 workers10.5–28.0 million3–8 hours a worker
13New Jersey53.0%2,228,870 workers12.7–33.8 million3–8 hours a worker
14Rhode Island52.7%245,440 workers1.4–3.7 million3–8 hours a worker
15Georgia52.7%2,486,650 workers14.0–37.3 million3–8 hours a worker
16Arizona52.5%1,666,800 workers9.4–25.0 million3–8 hours a worker
17North Carolina51.9%2,509,310 workers14.2–37.9 million3–8 hours a worker
18Delaware51.8%237,300 workers1.4–3.6 million3–8 hours a worker
19Vermont51.7%145,980 workers0.8–2.2 million3–8 hours a worker
20Illinois51.5%3,093,040 workers17.6–46.9 million3–8 hours a worker
21Minnesota51.4%1,484,180 workers8.5–22.6 million3–8 hours a worker
22California51.0%9,034,720 workers52–139 million3–8 hours a worker
23Alaska50.9%158,150 workers0.9–2.4 million3–8 hours a worker
24Oregon50.7%953,690 workers5.5–14.7 million3–8 hours a worker
25Pennsylvania50.6%3,027,180 workers17.4–46.4 million3–8 hours a worker
26Ohio50.6%2,772,380 workers15.7–42.0 million3–8 hours a worker
27Michigan50.5%2,184,440 workers12.5–33.3 million3–8 hours a worker
28Nebraska50.2%504,270 workers2.9–7.7 million3–8 hours a worker
29Missouri49.9%1,432,880 workers8.2–21.9 million3–8 hours a worker
30Tennessee49.3%1,578,090 workers9.1–24.4 million3–8 hours a worker
31Montana49.2%241,120 workers1.4–3.8 million3–8 hours a worker
32New Mexico49.2%417,830 workers2.4–6.5 million3–8 hours a worker
33Maine49.2%303,730 workers1.8–4.7 million3–8 hours a worker
34Oklahoma49.1%830,810 workers4.8–12.8 million3–8 hours a worker
35South Carolina49.0%1,112,870 workers6.5–17.2 million3–8 hours a worker
36Hawaii48.5%288,970 workers1.7–4.5 million3–7 hours a worker
37Idaho48.3%403,090 workers2.4–6.3 million3–8 hours a worker
38Kansas48.2%681,270 workers4.0–10.7 million3–8 hours a worker
39Wisconsin47.9%1,383,230 workers8.1–21.6 million3–7 hours a worker
40Arkansas47.8%609,210 workers3.6–9.5 million3–7 hours a worker
41West Virginia47.4%323,310 workers1.9–5.0 million3–7 hours a worker
42Iowa47.3%723,410 workers4.3–11.5 million3–8 hours a worker
43Kentucky47.2%926,650 workers5.4–14.5 million3–7 hours a worker
44Alabama46.9%965,880 workers5.7–15.2 million3–7 hours a worker
45Louisiana46.4%859,720 workers5.1–13.7 million3–7 hours a worker
46South Dakota46.2%207,610 workers1.2–3.3 million3–7 hours a worker
47North Dakota45.9%191,450 workers1.1–3.0 million3–7 hours a worker
48Nevada45.6%683,770 workers4.0–10.8 million3–7 hours a worker
49Indiana45.1%1,428,930 workers8.6–22.9 million3–7 hours a worker
50Wyoming45.0%121,130 workers0.7–2.0 million3–7 hours a worker
51Mississippi44.0%495,910 workers3.0–8.0 million3–7 hours a worker
not rankedGuamTerritory46.5%26,030 workers151,000–401,0003–7 hours a worker
not rankedPuerto RicoTerritory49.4%443,490 workers2.6–7.0 million3–8 hours a worker
not rankedVirgin IslandsTerritory48.8%13,100 workers74,000–199,0003–7 hours a worker

At least half of the work: an AI exposure index of 0.50 or more, the mean of our two published methods. An occupation’s own page calls 0.40–0.65 moderate exposure. Hours: a 40-hour week × the importance-weighted share of each occupation’s tasks AI does or helps with × the 15–40% time saved in published experiments, summed over every worker counted. Employment: BLS Occupational Employment and Wage Statistics, May 2025, state estimates; cells BLS suppresses are left out rather than counted as zero, so the sums cover 93–100% of each state’s published employment. Tasks: O*NET Database 31.0. Territories are listed apart and not ranked. Method

The least exposed jobs

Occupations with at least 50,000 workers where AI does or helps with the smallest share of the work. No job is AI-proof in every task — these are the ones whose work mostly needs hands, a place or a person present.

  1. 1Landscaping and Groundskeeping Workers0% · 0 of 12 tasks
  2. 2Industrial Truck and Tractor Operators0% · 0 of 9 tasks
  3. 3Dishwashers0% · 0 of 12 tasks
  4. 4Bus and Truck Mechanics and Diesel Engine Specialists0% · 0 of 12 tasks
  5. 5Helpers--Electricians0% · 0 of 12 tasks
  6. 6Maids and Housekeeping Cleaners3% · 1 of 12 tasks
  7. 7Cooks, Fast Food3% · 1 of 12 tasks
  8. 8Orderlies6% · 2 of 12 tasks
  9. 9Coating, Painting, and Spraying Machine Setters, Operators, and Tenders6% · 2 of 12 tasks
  10. 10Welders, Cutters, Solderers, and Brazers6% · 2 of 12 tasks

How to read this

This counts tasks, not jobs. A task AI does or helps with is language or information work a model can draft, sort or check. It is not a forecast that anyone will automate the job; for what the labor market expects, each occupation’s page shows the Bureau of Labor Statistics ten-year projection beside it.

The hours are a range, with a method. A 40-hour week × the importance-weighted share of the job’s O*NET tasks that AI does or helps with × the 15–40% time saved in published experiments: customer-support agents with an AI assistant (Brynjolfsson, Li & Raymond, 2023), professional writing tasks (Noy & Zhang, Science, 2023) and consultants at BCG (Dell’Acqua et al., 2023). An estimate for planning, not a promise.

New entrants feel it first. Employment of 22- to 25-year-olds in AI-exposed occupations is 19% below trend. Stanford Digital Economy Lab, August 2026.

Learn to do your job with AI

Short practice on the tasks of your own occupation, in the AI tool your employer allows, checked when you paste the result. The first two modules are free, and your résumé never leaves your browser.