Quick answer
No. In the Census Bureau's survey for September 2026, the median non-tech sector had 18% of businesses using AI, and 12 of 16 non-tech sectors had a 95% interval above a 10% line we set after seeing the data. Tech is still far ahead, about 2.8 times non-tech (2.0 to 3.7 times). This measures any AI use, not AI agents.
Key takeaways
- Use outside tech is substantial (median non-tech sector 18%, range 6.2% to 41%), and tech is still about 2.8 times non-tech. Both are true together (Census BTOS, share of employer businesses using AI in any business function).
- The 10% line was chosen after we had seen the first and last values, so the 'not only tech' result is not a blind test; at 5% 14 of 16 sectors qualify and at 20% only 4 of 16.
- No public dataset measures AI agents as such. These data measure any AI use (Census) and payments for AI products or services seen at one card company (Ramp).
Is AI only used by tech companies?
No. Many non-tech sectors use AI, but tech uses it about 2.8 times as much (2.0 to 3.7 times). The measure is the Census Bureau's survey of the share of employer businesses that used AI in any business function in the last two weeks, which includes chatbots and machine learning. No public dataset measures "AI agents" as such. In the final period, collected up to 20 September 2026, the median non-tech sector was at 18%, the range was 6.2% to 41%, and 12 of 16 non-tech sectors had a 95% interval above 10%.
We chose the 10% line, and the rule for counting it, after seeing the first and last values of the series, so this is not a blind test. At 5%, 14 of 16 non-tech sectors qualify, at 20% only 4 of 16 do, and the ratio of non-tech to tech use is 0.36 (95% CI 0.27 to 0.50); counting Information alone as tech gives 0.327, which our rule calls mixed. The solid conclusion is the narrower one: use outside tech is not negligible.
What did we measure, and how?
We compared AI use in tech sectors with every other sector, using two public sources. The first is the Census Bureau's Business Trends and Outlook Survey (BTOS): the share of employer businesses that used AI in any business function in the last two weeks, by industry sector, across biweekly periods from 17 November 2025 to 20 September 2026. The second is the Ramp AI Index, distributed by the Federal Reserve Bank of St. Louis: the share of one card company's customers with a paid AI subscription, by industry, monthly from January 2023 to August 2026.
We started from the assumption that AI agents are only used in tech. No public file measures agents, so we tested the nearest question: is AI use confined to tech sectors? The direction was visible before we froze the protocol, which is why the thresholds are disclosed and not presented as a confirmation. Here, "tech" means Information plus Professional, scientific and technical services. The usable sample is n = 384 sector-by-period figures, and Census suppressed another 34, which we never filled in.
Those 384 figures move together over time within each sector, so the result is less certain than the count suggests. Technical detail: the design effect is 18.9, which leaves an effective n = 20. A confidence interval ("95% CI") is the range the true figure very likely falls in, given that Census surveyed a sample and did not ask every business.
Analysis by Deivy Hernandez, from public data files.
How does AI use compare by sector?
Tech is at 44% and non-tech sectors average 16%, a gap of 28 percentage points (95% CI 22 to 35 percentage points). These are the Census figures for the last period. The group averages are weighted so that more precisely measured sectors count more, not by number of businesses. The tech average is mostly the professional, scientific and technical sector. Each figure describes a group of businesses, not a firm, and each carries a standard error that Census publishes.
| Sector (BTOS, latest period: share of employer businesses using AI in any business function) | Group | Share using AI |
|---|---|---|
| Information | Tech | 49.0% |
| Professional, scientific and technical services | Tech | 43.9% |
| Finance and insurance | Non-tech | 41.0% |
| Educational services | Non-tech | 30.0% |
| Real estate and rental | Non-tech | 29.6% |
| Health care and social assistance | Non-tech | 25.5% |
| Arts, entertainment and recreation | Non-tech | 22.6% |
| Administrative and support services | Non-tech | 20.8% |
| Manufacturing | Non-tech | 19.5% |
| Wholesale trade | Non-tech | 19.2% |
| Retail trade | Non-tech | 16.5% |
| Construction | Non-tech | 15.4% |
| Other services | Non-tech | 15.0% |
| Transportation and warehousing | Non-tech | 13.0% |
| Utilities | Non-tech | 11.8% |
| Accommodation and food services | Non-tech | 10.1% |
| Mining, oil and gas | Non-tech | 8.0% |
| Agriculture, forestry, fishing | Non-tech | 6.2% |
Management of companies has no published estimate for the latest period. Multi-sector or multi-state companies, reported separately and not part of the comparison, are at 36%. Agriculture, mining, utilities and management of companies are partly or wholly suppressed, so the study cannot speak fully for them. For a closer look at which industries sit where, see our earlier piece on AI use by business size and industry.
What does the Ramp card data add?
It shows higher levels of use, from a very different sample. Ramp measures the share of one card company's customers with an observed payment for an AI product or service. That is not a probability sample and it has no standard errors, so read these as descriptions of those customers, not estimates for US businesses. The six non-tech industries are construction, finance, health care, hospitality, manufacturing and retail.
| Ramp AI Index (one card company's customers with an observed payment for an AI product or service) | January 2023 | August 2026 |
|---|---|---|
| Information sector (the Ramp series FRED labels "Information") | 15% | 81% |
| Non-tech, mean of six industries (unweighted) | 4% | 50% |
In August 2026, 6 of 6 non-tech industries were above 10%. The table gives two endpoints only, and we draw no trend from them.
Is the gap between tech and other sectors changing?
In the Census data the gap in percentage points rose a little over about 10 months; the relative gap is inconclusive. The tech minus non-tech gap changed by 0.21 percentage points per period (95% CI 0.12 to 0.31). That rests on two tech sectors and the equivalent of about 20 independent data points, so it is a short window and a soft estimate. For the relative gap, the interval includes zero, so we cannot say whether it is stable. This says nothing about the longer run.
What can this data not show?
It cannot show AI agents, whether AI adds value, why sectors differ, or anything about a single state or region. The rest of the limits:
- Both sources are national, and neither measures whether AI adds value.
- BTOS covers employer businesses, not businesses without employees.
- Sectors 11, 21, 22 and 55 are partly or wholly suppressed.
- The 22 Census periods cover about 10 months, so any change is a short-window change.
- There are only two tech sectors and one Ramp Information-sector series, so the tech results describe those, not tech in general.
- Ramp customers are one vendor's non-probability sample, and Ramp industries do not map onto Census sectors.
How can you check it?
Both inputs are public. The Census file is the Sector workbook on the BTOS data page, with direct file at Sector.xlsx, question 7 (AI use). The Ramp series are on FRED, for example tech, retail and construction; the other four industries have matching series pages. Census asks that its figures be cited as the U.S. Census Bureau, Business Trends and Outlook Survey; FRED marks the Ramp series as requiring citation.
What can you do with this?
Do not assume AI is only a tech thing; look at your own sector's row in the table. Businesses in non-tech sectors report using AI, at levels from 6.2% to 41% depending on the sector. The data do not show whether it pays off. If you want a practical starting point, read what AI agents for small business can and cannot do before you buy anything.
FAQ
Does this tell me whether AI agents are used outside tech?
No. The Census question counts any use of AI in any business function, which includes chatbots and machine learning. The Ramp index counts businesses with an observed payment for an AI product or service, among one card company's customers. Neither separates out AI agents, and no public dataset we found does.
Which industries are counted as tech here?
Two Census sectors: Information, and Professional, scientific and technical services. Everything else with a published estimate is non-tech. The tech average is mostly the second sector, which has the more precise estimate. Information alone is at 49.0%.
Is the Census figure the share of my industry’s companies that use AI?
It is the share of employer businesses in a sector group that said yes, as a group figure. It does not tell you the chance that a particular firm uses AI, and it leaves out businesses with no employees.
Is the median the share of businesses in my industry?
No. The median is the middle value across 16 non-tech sectors. To see your own industry, find its row in the table: each row is the share of employer businesses in that sector group that said yes, and it describes a group, not any single firm.
Why is finance so high if it is not tech?
The survey shows Finance and insurance at 41.0%, the highest non-tech sector, and it does not say why. The data measure how many businesses report using AI, not the reasons behind it.