Opens in a new tab
needsahuman
Search ⌘KGet the data
Every job scored on three questions, with the evidence grade and the range

Will AI take your job? Measured, not guessed.

Every job scored on what AI can do today, whether it does it better than a person, and when it could take over. The answer is one word, and the working is underneath.

Will AI replaceAccountants and Auditors?See the job →
A little.Still needs a human: 66/100 ↑ safer
10% AI does it67% AI helps23% needs a human
Each figure is a job. Amber is the work that still needs a human; grey is what AI helps with; white is what AI can do.
← AI can do the jobStill needs a human · 0 to 100 · higher is saferStill needs a human →
020406080100
Largely.
0 jobs · already happening
0–19
Mostly.
0 jobs · smaller job by the 2030s
20–39
Partly.
36 jobs · task erosion, fewer juniors
40–59
A little.
562 jobs · AI helps, the job stays
60–79
Nah.
325 jobs · AI can't do it yet
80–100 · the good end
923
Occupations scored so far
O*NET 31.0 · 2026-Q4
5%
Of US workers in scored jobs where AI covers over half the task time
OEWS employment × coverage
0
Jobs with direct, blind-graded evidence (grade A)
GDPval, Vals, MAI-DxO
2045
Median replacement year across scored jobs. 80% of scored jobs between 2038 and 2057.
10,000 scenarios per job

How the work could shift

The share of US workers whose job would sit in each band, from today to 2060, as AI reaches each job at the pace our timeline model expects.

Today
35%
of US workers are in jobs that mostly need a person (our top band, Nah.)
0% are in jobs AI could largely do (Largely.)
By 2045
5%
of US workers would still be in jobs that mostly need a person (our top band, Nah.)
58% would be in jobs AI could largely do (Largely.)
By 2060
5%
of US workers would still be in jobs that mostly need a person (our top band, Nah.)
92% would be in jobs AI could largely do (Largely.)

Each job's score moves towards the bottom band (Largely: AI could largely do the job) by the year it could be largely automated, slowly at first and faster later, as adoption usually goes. Each job is run as ten scenarios across its replacement range, so every figure mixes early and late outcomes. The model stops at 2060. How the timeline works

Share of US workers in each verdict band, today to 20600%25%50%75%100%2026: Nah. 34.9% of US workers35%2026: A little. 59.4% of US workers59%2026: Partly. 5.6% of US workers6%Today2030: Nah. 17.5% of US workers18%2030: A little. 51.4% of US workers51%2030: Partly. 28.9% of US workers29%2030: Mostly. 2.2% of US workers20302035: Nah. 5.3% of US workers2035: A little. 29.9% of US workers30%2035: Partly. 29.8% of US workers30%2035: Mostly. 20.3% of US workers20%2035: Largely. 14.7% of US workers15%20352040: Nah. 5.1% of US workers2040: A little. 7.5% of US workers8%2040: Partly. 24.4% of US workers24%2040: Mostly. 28.0% of US workers28%2040: Largely. 35.0% of US workers35%20402045: Nah. 5.1% of US workers2045: A little. 1.9% of US workers2045: Partly. 10.4% of US workers10%2045: Mostly. 24.5% of US workers25%2045: Largely. 58.1% of US workers58%20452050: Nah. 5.1% of US workers2050: A little. 1.8% of US workers2050: Partly. 1.6% of US workers2050: Mostly. 17.7% of US workers18%2050: Largely. 73.8% of US workers74%20502055: Nah. 5.1% of US workers2055: A little. 1.8% of US workers2055: Mostly. 6.5% of US workers7%2055: Largely. 86.6% of US workers87%20552060: Nah. 5.1% of US workers2060: A little. 1.8% of US workers2060: Mostly. 1.6% of US workers2060: Largely. 91.5% of US workers92%2060
Will AI replace the job?Largely.Mostly.Partly.A little.Nah.
Share of US workers whose job would sit in each band, year by year. Updated with every release.
Show the data
YearLargelyMostlyPartlyA littleNah
Today (2026)0.0%0.0%5.6%59.4%34.9%
20300.0%2.2%28.9%51.4%17.5%
203514.7%20.3%29.8%29.9%5.3%
204035.0%28.0%24.4%7.5%5.1%
204558.1%24.5%10.4%1.9%5.1%
205073.8%17.7%1.6%1.8%5.1%
205586.6%6.5%0.0%1.8%5.1%
206091.5%1.6%0.0%1.8%5.1%

How we score. Three questions, one word.

Open data, versioned scores, a public method. Commercial relationships never change a score.

Read the methodology

Can AI do it?

Share of the job's task time AI can handle today, from O*NET tasks, the Anthropic Economic Index and Microsoft's usage data. The amber slice is what still needs a person.

Is it better than a person?

Measured against real professionals where a blind test exists, and graded A to D for how good the evidence is. Grade D says “not yet measured”, never a guess.

When could it be replaced?

10,000 scenarios per job combining capability growth, adoption and friction. Always a range, never a date.

THE NAH REPORT · EVERY QUARTER

What moved this quarter. Jobs that changed their word.

Each release we rescore every job, publish what moved and why, and keep the old numbers so anyone can check.

See what changedPress kit
OPEN DATA · CC BY 4.0

Every score, every release, one CSV.

Three scores per job, evidence grades, ranges and history. Cite it, chart it, build on it. Attribution required, that's all.

Download CSVJSON APIChangelog