The Revenue AI Report

Adoption is real, measurable, and slower than the discourse

US government data has tracked firm-level AI use every two weeks for three years. It says 22.4%. A payments dataset says 55.73%. Both are right.

The short answer

What does the research show about Adoption is real, measurable, and slower than the discourse?

US government data has tracked firm-level AI use every two weeks for three years. It says 22.4%. A payments dataset says 55.73%. Both are right. The Census Bureau Business Trends and Outlook Survey publishes firm-level AI use biweekly with standard errors, broken out by employment size and 18 sectors. It is free, it now runs 74 periods, and almost nobody in the GTM conversation charts it.

Evidence

  • Median AI spend per employee per month on Ramp went from $2.32 to $11.95 between January 2023 and July 2026.
  • Eurostat, EU27 enterprises with 10 or more employees using at least one AI technology: 7.65% in 2021, 8.06% in 2023, 13.48% in 2024, 19.95% in 2025.
  • Indeed Hiring Lab: US job postings mentioning AI rose from 1.71% to 6.17% across 91 monthly points to July 2026.

Supporting pages

The Census Bureau Business Trends and Outlook Survey publishes firm-level AI use biweekly with standard errors, broken out by employment size and 18 sectors. It is free, it now runs 74 periods, and almost nobody in the GTM conversation charts it.

Two things about the series must be shown rather than smoothed. The question wording changed on November 17, 2025 and the measured rate jumped from 10.0% to 17.3% at the break. Periods 85 to 87 are null because of the October 1 to November 13, 2025 federal funding lapse. Any single continuous line across that break publishes a 7-point artifact as growth.

What this page is

Three years of firm-level AI adoption from a free government series, read against private datasets that measure a different universe and report a rate more than double it.

The argument

Adoption is real and it is slower than the discourse. The spread between credible adoption numbers is caused by who is counted and how the question is worded, not by anyone being wrong.

How to read it

  • The Census series has a documented break. Question wording changed on 17 November 2025 and the rate moved from 10.0 to 17.3 percent at that point. A single continuous line across the break publishes a 7-point artifact as growth.
  • Periods 85 to 87 are null because of the 1 October to 13 November 2025 federal funding lapse. Interpolating them invents data.
  • The 55.73 percent payments figure and the 22.4 percent Census figure count different populations. Neither is the adoption rate.

US firms using AI went from 3.7% to 22.4% in three years. One of those points is a wording change.

Selected periods from the biweekly series. Standard errors are published for every period.

09182635question wording changed November 17, 202522.425.9Sep 2023Feb 2024Aug 2024Feb 2026Apr 2026Jul 2026Aug 2026Share of US firms, %
  • Using AI now
  • Expect to use AI in six months

What this does not say

The jump between two of these points is partly a change in how the question was worded. It is not entirely a change in behavior.

Publisher
US Census Bureau, Business Trends and Outlook Survey
Sample and method
Periods 31 to 107, standard errors published
Field dates
September 2023 to August 2026

Wording break at November 17, 2025 and a collection gap during the federal funding lapse.

High confidence

One dataset says 56% of businesses use AI. Another says 22%. Both are right.

Ramp measures paid transactions at AI vendors. Census asks firms whether they used AI to produce goods or services.

01633496555.7321.7Jan 2023Jan 2024Jan 2025Jan 2026Jul 2026Share of businesses, %
  • Ramp, paid AI vendor penetration
  • Census BTOS, AI use now

What this does not say

Neither number is wrong. They count different populations, and no reconciliation between them exists.

Publisher
Ramp Economics Lab AI Index and US Census Bureau BTOS
Sample and method
Ramp: more than 50,000 US businesses with at least one paid AI transaction in the month, 43 monthly points. Census: national estimate. Intermediate Census points are interpolated between published periods for shape.
Field dates
Not published by the source.

High confidence in each series on its own, low confidence in the comparison. Ramp card and bill-pay customers are not a representative sample of US firms.

Low confidence

Also in the record

Figures that sit alongside these charts.

  • Median AI spend per employee per month on Ramp went from $2.32 to $11.95 between January 2023 and July 2026.
  • Eurostat, EU27 enterprises with 10 or more employees using at least one AI technology: 7.65% in 2021, 8.06% in 2023, 13.48% in 2024, 19.95% in 2025.
  • Indeed Hiring Lab: US job postings mentioning AI rose from 1.71% to 6.17% across 91 monthly points to July 2026.
  • The St. Louis Fed published directly on the measurement problem: how you ask changes what firms report.

The brief

What is going on here, and why it matters.

The charts above are the evidence. This is the read: what the data describes, the mechanism behind it, where the argument could be wrong, and what a revenue team does about it.

01

What is actually happening

Firm-level AI use in the United States moved from 3.7 percent to 22.4 percent across three years of biweekly measurement with published standard errors, sector breakouts, and employment-size cuts. That is one of the best-instrumented technology adoption series available, and it is free. It is also nearly absent from the GTM conversation, which prefers vendor surveys with larger numbers and no standard errors.

The EU series moves the same direction on a different clock: 7.65 percent in 2021, 8.06 percent in 2023, 13.48 percent in 2024, 19.95 percent in 2025. Job postings mentioning AI rose from 1.71 to 6.17 percent across 91 monthly points. Three independent instruments, three consistent slopes, all well under the numbers used in board decks.

02

Why the numbers disagree

A payments dataset sees firms that transact on a card platform, which skews toward digitally native businesses that were always going to adopt first. A government survey samples the whole business population, including the long tail of firms with no software budget. The 33-point spread is a sampling frame difference, and the St. Louis Fed published directly on the adjacent problem: how you ask changes what firms report.

Spend data adds the third view. Median AI spend per employee per month on Ramp went from $2.32 to $11.95 between January 2023 and July 2026. That is a fivefold rise on a small base. Both facts belong in the same sentence, and usually only the multiple travels.

03

What this means for a revenue plan

If you sell AI-adjacent software to the broad market, roughly four in five firms are still not using AI at all by the strictest available measure. The market is early, not saturated, and messaging that assumes an informed buyer is talking to a fifth of the room.

If you sell into digitally native segments, the payments figure is closer to your reality and your differentiation problem is already severe. The correct move is to name your segment's adoption rate rather than the market's, because the two are more than two times apart.

04

Where this argument could be wrong

Firm-level use counts a firm, not a workforce. A 10,000-person enterprise where one team uses AI counts the same as a five-person shop where everyone does. Employee-level adoption is plausibly far higher than 22.4 percent, and this series cannot see it.

Self-reported use also depends on whether respondents count AI features embedded in software they already bought. As those features become default, the series will rise for reasons that have nothing to do with a decision anyone made.

What to do with it

The move, by seat.

CMO
Quote the adoption rate for your segment and name the sampling frame in the same line. A single market-wide number is not defensible in either direction.
Strategy
Model the wording break explicitly. Any trend line you build across 17 November 2025 overstates growth by roughly 7 points.
Founders
Treat the long tail as unadopted rather than late. Four in five firms are a category education problem, not a competitive displacement problem.

Questions this page answers

What the data says, in plain language.

What does the research show about Adoption is real, measurable, and slower than the discourse?
US government data has tracked firm-level AI use every two weeks for three years. It says 22.4%. A payments dataset says 55.73%. Both are right. The Census Bureau Business Trends and Outlook Survey publishes firm-level AI use biweekly with standard errors, broken out by employment size and 18 sectors. It is free, it now runs 74 periods, and almost nobody in the GTM conversation charts it.
What does the figure "US firms using AI went from 3.7% to 22.4% in three years. One of those points is a wording change" show?
Selected periods from the biweekly series. Standard errors are published for every period. Source: US Census Bureau, Business Trends and Outlook Survey. Periods 31 to 107, standard errors published Fielded September 2023 to August 2026. Confidence: High. Caveat: Wording break at November 17, 2025 and a collection gap during the federal funding lapse.
What does the figure "One dataset says 56% of businesses use AI. Another says 22%. Both are right" show?
Ramp measures paid transactions at AI vendors. Census asks firms whether they used AI to produce goods or services. Source: Ramp Economics Lab AI Index and US Census Bureau BTOS. Ramp: more than 50,000 US businesses with at least one paid AI transaction in the month, 43 monthly points. Census: national estimate. Intermediate Census points are interpolated between published periods for shape. Confidence: Low. Caveat: High confidence in each series on its own, low confidence in the comparison. Ramp card and bill-pay customers are not a representative sample of US firms.
What else sits alongside these figures?
Median AI spend per employee per month on Ramp went from $2.32 to $11.95 between January 2023 and July 2026. Eurostat, EU27 enterprises with 10 or more employees using at least one AI technology: 7.65% in 2021, 8.06% in 2023, 13.48% in 2024, 19.95% in 2025. Indeed Hiring Lab: US job postings mentioning AI rose from 1.71% to 6.17% across 91 monthly points to July 2026. The St. Louis Fed published directly on the measurement problem: how you ask changes what firms report.
Where does this data come from?
Every figure is reproduced from a named publisher: US Census Bureau, Business Trends and Outlook Survey, Ramp Economics Lab, AI Index, Eurostat ICT enterprise survey, St. Louis Fed, Measuring AI adoption by firms: how you ask matters, Indeed Hiring Lab AI tracker. Sample, field date, and confidence are shown on each chart. Sources marked as vendor research are labelled on the page.

Cite this page

Permanent URL and suggested citation.

https://www.therevenueaireport.com/research/adoption-curve

Kvarfordt, Jonathan. "Adoption is real, measurable, and slower than the discourse." The Revenue AI Report, Research Library. https://www.therevenueaireport.com/research/adoption-curve

Figures on this page are reproduced from the publishers listed below. Cite the original publisher for the underlying data, and this page for the compilation and framing.

Sources

Every publisher used on this page.

If a metric, model term, or method on this page is unfamiliar, every one of them is defined in The AI and Revenue Dictionary. Sample size, field date, and confidence tags are explained there too.

Could not confirm

What we looked for and did not find.

Claims found during research and not charted

Nothing on this theme was dropped for sourcing. Every claim we found that met the standards on the Research hub is charted above, and anything that failed them would be listed here by name.

Read the analysis

Issues built on this theme.

Research on this site is the evidence layer. These essays take the numbers above and apply them to real decisions, so you can see how the data reads in practice.

How to cite this research

Written by Jonathan Kvarfordt, Founder and Principal Analyst, The Revenue AI Report. Published under CC BY 4.0.

APA

Kvarfordt, J. (2026). Adoption is real, measurable, and slower than the discourse. The Revenue AI Report. Retrieved from https://www.therevenueaireport.com/research/adoption-curve

MLA

Kvarfordt, Jonathan. "Adoption is real, measurable, and slower than the discourse." The Revenue AI Report, 31 Aug. 2026, www.therevenueaireport.com/research/adoption-curve.

BibTeX

@misc{kvarfordt2026adoptioncurve,
  author = {Kvarfordt, Jonathan},
  title = {Adoption is real, measurable, and slower than the discourse},
  year = {2026},
  publisher = {The Revenue AI Report},
  url = {https://www.therevenueaireport.com/research/adoption-curve}
}

Share this research

Posting to Instagram or TikTok? Copy the link, it carries the title, summary and share image.

Next theme

Rollback is a measured pattern, not an anecdote

74% of enterprises have pulled a deployed AI agent over a governance failure. The rate among self-described mature guardrails is higher, not lower.

Subscribe

Get the weekly issue built on this data.

Arrives weekly by email. Free. Unsubscribe anytime. By subscribing you agree to our Privacy policy and Terms. We never sell or share the list.