AI Jobs, AI Adoption and AI Productivity: what the data shows

56.8% of US businesses pay for AI tools, up from 7.5% in January 2023 and 43.5% a year ago 5.3% of GDP in tech investment ($1.77 trillion a year), above the dot-com peak of 4.5% -381k information-sector jobs since November 2022 (-12.2%), while total payrolls added 5.1 million 6.6% unemployment for bachelor's graduates aged 20-24, against 4.3% for all workers Spending and productivity have accelerated, total employment has held, and the strain is concentrated at white-collar entry points. The graduate gap is 2.3 points; it averaged 0.3 over 2001-2019.

Twelve indicators, each shown at its latest reading, a year ago and in November 2022.

Six of the deck's core series at a glance.

All 56.8%, large 66.7%, medium 63.2%, small 51.2% (October 2026).

Up 13.3 points in a year. Large firms lead at 67%.

All businesses +13.3 points in the past year against +10.1 the year before; large firms +14.9.

At work 40.5% in Q3 2026 (33.3% in Q3 2024, peak 45.2%); any use 60.8%.

Workplace use slipped from 45.2% in Q2: one quarter, worth watching.

Management 5.2% of hours, all workers 2.1%, food preparation 0.7%.

7x gap between the most and least affected groups.

Manufacturing +2.6 points in a year, all workers +0.3, arts, design and media -0.6.

Creative and teaching roles report less time saved than a year ago.

5.35% of GDP in Q3 2026 against 4.46% at the 2000 peak; computers 1.53% against 1.05%.

A record 5.35% of GDP, $1.77 trillion a year.

Computers +70.5% y/y to a $504bn annual rate, software +9.0%, all business investment +12.1%.

Computer investment is growing 8x as fast as software.

Taiwan ICT $393bn, 6.0x the year to November 2022; Korea chips $307bn, 2.6x.

$393bn in 12 months, up from $65bn in the year to November 2022.

ICT is 44% of exports (was 13% in 2022); the US takes 32% (was 16%).

ICT share has more than tripled since 2022.

Korea SSD exports $42bn (2.8x late 2022); Taiwan components $307bn (1.5x).

Virginia commercial load +65% since 2019, Texas +31%, all US electricity +8%.

Northern Virginia is the largest US data center cluster.

Since Q4 2022: real GDP +10.5%, output per hour +10.6%, payrolls +3.1%.

Output up 10.5% with payrolls up about 3%: the gap is productivity.

2.7% a year since Q4 2022 against 1.2% across 2010-2019. The timing fits AI; the cause is not proven.

Timing fits AI. Rates, mix shifts and the post-pandemic rebound also contribute.

Profits 13.1% of GDP (from 10.7% in Q4 2022); labor share index 93.7 (from 96.9).

Since November 2022: information -12.2%, computer systems design -5.2%, construction +6.4%, all payrolls +3.3%.

Information has lost 381k jobs while construction added 504k.

Health care +3,074k and government +953k lead; temporary help -511k and information -381k trail.

Health care alone added 61% of all net new jobs.

Correlation -0.73 across 7 sectors: information saves 4.4% of hours and is -12.2% on jobs; health care 1.4% and +14.7%.

Correlation -0.73 across seven sectors. An association, not proof of cause.

Since November 2022: architecture +9.6%, legal +6.0%, finance -1.1%, computer -5.2%.

Openings 7.5 million against a 12.3 million peak in 2022; temporary help 2.53 million against 3.16 million.

6.6% against 4.3% in October 2026; in November 2022 it was 5.3% against 3.7%.

Before 2018 young graduates tracked the average. The gap is now 2.3 points.

Gap 2.3 points in October 2026; it has stayed above 1.5 since 2023 and averaged 0.3 over 2001-2019.

The gap began widening around 2014, well before generative AI.

782k people are unemployed and seeking a first job, up from 482k; 10.7% of the unemployed, from 8.0%.

Bachelor's and higher 2.0% to 2.8% (+38%); high school 4.0% to 4.4% (+8%).

Rates are still lowest for the most educated; the rise is proportionally largest for them.

Jobless-rate advantage of a degree: 1.6 points now against a 2.8-point average over 1993-2019.

Near the 2019 low: a degree's edge is about 56% of its long- run average.

Average spell 25.1 weeks (from 23.3 in November 2022); long-term share 26.0% (from 23.0%); U-6 underemployment 8.0% from 6.9%.

Bank views cited from J.P. Morgan US Market Intelligence (November 12, 2026) and Morgan Stanley Research (November 11, 2026). The figures are the banks' own.

Three qualitative paths, each tied to the FinzData series that would confirm it. No forecasts are implied.

Cause: The timing fits AI, but higher interest rates and the unwinding of 2021-2022 tech over-hiring explain part of the weakness in information and computer services jobs. Measurement: Business adoption comes from corporate card spending and worker adoption from a survey. Both can move with sample changes, as the Q3 2026 dip in workplace use shows. Graduates: The graduate gap began widening before 2022. AI may be extending a trend that started with degree oversupply and slower professional hiring. Revisions: Payrolls, productivity and GDP are revised. Every figure here is the latest published vintage as of November 12, 2026.

Economy: Growth is leaning on investment and productivity, not hiring. Tech investment is 5.3% of GDP and output per hour is growing 2.7% a year. A capex pause would hit GDP and Asian exports before US jobs. Employment: Total employment is holding and the mix is moving. Information has lost 381k jobs since November 2022 while health care and construction keep adding. Entry level: Fewer first rungs. 782k people are unemployed and seeking a first job, and the tasks that trained juniors are the ones AI saves the most time on. Graduates: A degree still lowers unemployment, but its advantage is 1.6 points against a 2.8-point long-run average. Field of study and hands-on experience matter more. Policy: The questions ahead are about entry-level pathways, retraining and who captures the productivity gain. Analysis, not investment advice.

Dated catalysts over the next four weeks and why each matters for the AI economy.

Vintages: Each observation is the latest published vintage in the FinzData macro store as of November 12, 2026. Point-in-time history is kept for every series. Averages: Unemployment rates by education and age are not seasonally adjusted, so they are shown as 12-month averages. Exports and electricity are rolling 12-month sums. Missing month: The October 2025 household survey was not published. That month is interpolated before averaging. Base date: November 2022 is used as the base for employment comparisons; Q4 2022 for quarterly series. Bank research: J.P. Morgan and Morgan Stanley views appear on one slide, cited, and are not FinzData data.

Every series in the deck is available point in time from FinzData.