Where AI is changing the American economy, and where it isn't yet.
Economic intelligence on adoption, jobs, wages, output, compute and communities across all twenty U.S. sectors. Every number is sourced, dated and labelled for what it is, so the story still holds if the evidence disagrees with it.
Three numbers, three denominators: all employer businesses, businesses that use one card platform, and conversations on one AI product. They measure different things and are never merged into one line on this site.
The index at a glance
One in four American businesses now uses AI. Where they are matters.
Businesses using AI, by state
Share of employer businesses, latest Census cycle.
State map. Highest: WY 40.1%, DC 37.2%, CO 31.2%.
Caveat. State samples are smaller than the national sample; close states should not be ranked.
Inspect the evidence
Text description. State map. Highest: WY 40.1%, DC 37.2%, CO 31.2%.
National AI use, two questions kept apart
Census changed the question in Nov 2025; shading marks a period with no collection.
Original question 3.7% to 10%; broadened question 17.3% to 24.2%.
Caveat. The jump in Nov 2025 mixes real growth with a wording change.
Inspect the evidence
Text description. Original question 3.7% to 10%; broadened question 17.3% to 24.2%.
The thesis, tested
AI becomes a broad growth factor only if every link holds.
Capability and cost must turn into adoption, adoption into new ways of working, those into more output, and output into higher real incomes and better services. Here is how strong the evidence is at each link today.
Training compute and data-center capacity are measured (Epoch, Census). Token prices on Ramp fell 14.8% in a month.
Two independent measures rise: Census (representative) and Ramp (vendor). Levels differ by definition.
Platform telemetry shows which tasks people bring to AI. It does not show how work is reorganized.
One firm-level study links heavy AI spending to headcount growth, mostly in Information. No sector output is attributable yet.
No assembled evidence yet ties AI to real wages, consumer prices or service quality at scale.
Evidence strength is an editorial rating of measurement quality, not of whether the effect is positive. A well-measured null is strong evidence.
Main Street, not just Wall Street
Payroll jobs since January 2019, with two adoption measures beside them.
Employment: BLS CES, seasonally adjusted, through Oct 2026 (latest two months preliminary). Census BTOS late Oct 2026; Ramp Oct 2026. Job changes have many causes; none of these lines isolates AI. All 20 sectors.
Job accounting
How many jobs has AI created?
Not currently established by the evidence assembled.
What exists are answers to different questions: how many people work in the industry that houses data centers, how many factory jobs a year of data-center spending could support, how fast heavy AI adopters hired. They cannot be added together.
See the reconciliationExisting employment in the industry that houses data centers. Not jobs created by AI.
Modeled employment supported at an investment level. Not new hires, not net.
A percentage effect within a selected sample. Do not multiply by national employment.
Shares of businesses, not numbers of jobs.
The report