AI use in US small businesses, by industry and size
One in four US employer firms now uses AI. Our analysis of Census Bureau survey data shows industry matters more than size, and where small firms still trail large ones.
24.6 %
of responding US employer firms used AI in the last two weeks (survey collected Oct 2026)
5 x
gap in AI use between the highest and lowest industry: information versus restaurants and hotels
3.5 x
how much more often firms with 250+ employees use AI in finance and accounting than firms with 1 to 4
Key findings
- Use is rising fast: 24.6% of responding firms used AI in the survey collected October 5 to 18, 2026, up from 17.3% in November 2025 under the same question.
- Industry matters more than size: across 15 industries, use ranges from 9.4% (accommodation and food services) to 46.7% (information), a 37-point spread, against 20 points across firm sizes.
- The smallest firms keep pace: firms with 1 to 4 employees use AI at 23.3%, in line with firms of 5 to 49 employees (21.1% to 23.8%); the step up comes at 100 employees and above.
- Late adopters are catching up fastest: construction rose 71% (9.2% to 15.7%) and manufacturing 69% (12.4% to 21.0%) from winter 2025-26 to late summer 2026, against 24% in information.
- AI mostly supports staff: among firms using AI, 43.7% used it to help with tasks employees do, 10.1% had it take over a task, and 95.7% reported no change in total employment.
- Small firms trail most in the back office: 5.9% of firms with 1 to 4 employees used AI in finance and accounting, against 20.5% of firms with 250 or more.
How many US businesses use AI today?
About one in four. In the survey collected October 5 to 18, 2026, 24.6% of responding employer firms said they used AI in some business function in the previous two weeks [1].
That is up from 17.3% in late November 2025: a rise of 7.3 points in eleven months, under the same question.
The Bureau’s original, narrower question tells the same story over a longer span. Use rose from 3.7% in September 2023 to 10.0% in September 2025, 2.7 times higher.
| Survey collected | Wording | Used AI | Expect to use |
|---|---|---|---|
| Sep 11 to 24, 2023 | Original | 3.7% | 6.3% |
| Mar 11 to 24, 2024 | Original | 4.4% | 6.7% |
| Sep 9 to 22, 2024 | Original | 5.9% | 8.7% |
| Mar 10 to 23, 2025 | Original | 7.8% | 10.5% |
| Sep 22 to Oct 5, 2025 | Original | 10.0% | 14.0% |
| Nov 17 to 30, 2025 | Current | 17.3% | 21.1% |
| Mar 9 to 22, 2026 | Current | 18.9% | 22.3% |
| Jun 15 to 28, 2026 | Current | 20.6% | 23.7% |
| Oct 5 to 18, 2026 | Current | 24.6% | 28.6% |
Plans run a little ahead of use. In October 2026, 28.6% expected to use AI in the next six months, 4.0 points above current use.
A large group is undecided. 25.8% said they did not know whether they would use AI in the next six months, slightly more than the share that uses it now.
Which industries use AI most, and which are catching up?
Industry is the strongest signal in the data. Over the latest six periods, 46.7% of information firms used AI, against 9.4% in accommodation and food services: a fivefold gap.
Four industries are at 30% or more: information, professional and technical services, finance and insurance, and educational services. Three are below 15%: other services (such as repair shops and salons), transportation and warehousing, and restaurants and hotels.
| Industry | Nov 2025 to Feb 2026 | Jul to Oct 2026 | Change, points | Expect to use |
|---|---|---|---|---|
| Information | 37.7% | 46.7% | +9.0 | 50.5% |
| Professional, scientific and technical services | 34.2% | 42.5% | +8.3 | 47.5% |
| Finance and insurance | 30.4% | 39.2% | +8.8 | 45.9% |
| Educational services | 30.6% | 34.9% | +4.3 | 39.4% |
| Real estate and rental and leasing | 23.5% | 29.7% | +6.2 | 34.9% |
| Health care and social assistance | 20.7% | 25.0% | +4.3 | 29.8% |
| Administrative and support services | 16.6% | 21.3% | +4.7 | 25.4% |
| Manufacturing | 12.4% | 21.0% | +8.6 | 24.9% |
| Wholesale trade | 13.6% | 19.9% | +6.3 | 24.1% |
| Arts, entertainment and recreation | 17.5% | 19.3% | +1.8 | 22.8% |
| Retail trade | 12.6% | 16.3% | +3.7 | 19.7% |
| Construction | 9.2% | 15.7% | +6.5 | 18.1% |
| Other services (repair, personal care and more) | 9.9% | 13.3% | +3.4 | 15.5% |
| Transportation and warehousing | 7.6% | 11.0% | +3.4 | 14.1% |
| Accommodation and food services | 8.0% | 9.4% | +1.4 | 12.6% |
The industries that started lowest are growing fastest. Construction rose 71%, from 9.2% to 15.7%, and manufacturing 69%, from 12.4% to 21.0%.
Wholesale trade (46%) and transportation and warehousing (45%) also grew faster than the national average of 30%. Information and professional services each grew 24%, from a much higher base.
Restaurants and hotels barely moved, from 8.0% to 9.4%. So did arts, entertainment and recreation, from 17.5% to 19.3%.
Does firm size matter as much as industry?
Less than industry does. Across seven size classes, use ranges from 21.1% to 40.7%, a 20-point spread, about half the 37-point spread across industries.
Firms with 1 to 4 employees use AI at 23.3%, in line with firms of 5 to 49 employees. The step up comes at 100 employees: 33.5% for 100 to 249 and 40.7% for 250 or more.
Source: AIMOCS analysis of US Census Bureau BTOS; average of six biweekly periods collected Jul 27 to Oct 18, 2026 (Employment Size Class and National files).
Industry wins even within a size band. A professional services firm with 1 to 4 employees (41.3%) uses AI at a higher rate than the largest firms in manufacturing (36.2%), health care (32.9%) or restaurants and hotels (21.7%).
| Industry | 1 to 4 | 5 to 9 | 10 to 19 | 20 to 49 | 250 or more |
|---|---|---|---|---|---|
| Information | 44.4% | 46.9% | 53.0% | 51.6% | 73.9% |
| Professional, scientific and technical services | 41.3% | 44.0% | 45.8% | 47.4% | 58.2% |
| Finance and insurance | 36.6% | 44.3% | 50.8% | 40.5% | 58.6% |
| Educational services | 37.0% | 30.0% | 29.8% | 35.8% | 53.9% |
| Real estate and rental and leasing | 29.8% | 27.0% | 30.3% | 31.0% | 61.4% |
| Health care and social assistance | 24.8% | 23.7% | 25.3% | 26.2% | 32.9% |
| Administrative and support services | 20.3% | 18.6% | 23.5% | 28.2% | 35.8% |
| Manufacturing | 19.6% | 19.9% | 19.4% | 20.9% | 36.2% |
| Wholesale trade | 17.8% | 19.5% | 22.3% | 25.9% | 41.1% |
| Arts, entertainment and recreation | 17.6% | 22.9% | 23.2% | 19.8% | 47.7% |
| Retail trade | 15.6% | 14.7% | 17.0% | 20.8% | 38.4% |
| Construction | 14.1% | 15.4% | 18.3% | 23.1% | 46.1% |
| Other services (repair, personal care and more) | 12.0% | 12.4% | 19.6% | 22.7% | n/a |
| Transportation and warehousing | 8.9% | 12.7% | 17.1% | 17.0% | n/a |
| Accommodation and food services | 8.6% | 7.5% | 9.6% | 10.6% | 21.7% |
Because most US employers are small, the national figure mostly describes small firms. In 2022, 89.4% of US employer firms had fewer than 20 employees [2].
Large firms employ most workers, so the picture looks different by headcount. We estimate that about 35% of employer-firm workers work at a firm that used AI in the last two weeks, against 23.3% of firms.
What do businesses use AI for?
The Bureau’s 2026 supplement asked which business functions firms used AI in over the previous six months [3]. Sales and marketing led at 14.3% of all firms, followed by strategy (12.4%), IT (11.4%) and research and development (11.2%).
| Business function | All firms | 1 to 4 employees | 250 or more | Large to small |
|---|---|---|---|---|
| Sales and marketing | 14.3% | 13.1% | 19.0% | 1.5x |
| Strategy and business development | 12.4% | 12.0% | 18.2% | 1.5x |
| Information technology | 11.4% | 11.2% | 25.9% | 2.3x |
| Research and development | 11.2% | 11.9% | 11.9% | 1.0x |
| Public relations and communication | 9.3% | 8.6% | 12.4% | 1.4x |
| Management and administration | 8.1% | 6.9% | 16.3% | 2.4x |
| Customer service | 7.6% | 7.1% | 14.2% | 2.0x |
| Legal and compliance | 7.1% | 6.8% | 10.5% | 1.5x |
| Finance and accounting | 6.6% | 5.9% | 20.5% | 3.5x |
| Human resources | 4.2% | 2.9% | 13.6% | 4.7x |
Small and large firms are close in sales, marketing and research. The gap is widest in the back office: finance and accounting (5.9% against 20.5%), human resources (2.9% against 13.6%) and management and administration (6.9% against 16.3%).
Where employees use generative AI, the work is mostly writing. Of those firms, 85.4% used it for writing or editing documents and emails, 49.9% to search for information and 44.6% to summarize or analyze documents.
Base: firms whose employees used generative AI in the last six months (20.8% of all firms). Firms could choose more than one task. Source: AIMOCS analysis of US Census Bureau BTOS AI supplement, about 117,000 responses collected Nov 17, 2025 to Feb 8, 2026.
Among firms that plan to use AI, the most common plans are sales and marketing (62.8%), strategy (57.4%) and research and development (48.3%). Customer service (40.0%) and finance and accounting (36.5%) follow.
Plans for finance and accounting rise with size, from 33.8% of planning firms with 1 to 4 employees to 60.1% of those with 250 or more.
Is AI replacing staff at small businesses?
So far, AI mostly helps the people already there. Among firms that used AI, 43.7% used it to support or improve a task an employee does.
Fewer used it to take over work: 10.1% had AI perform a task an employee used to do, and 10.6% used it for a task nobody did before.
Where AI did take over tasks, 70.9% of those firms said it was a small number. 7.1% said a large number.
Total employment barely moved. 95.7% of AI-using firms reported no change from AI in six months, 2.3% an increase and 2.0% a decrease.
The newest core question, first asked October 19 to November 1, 2026, points the same way. 87.3% of all firms said AI had not changed their headcount in six months; 6.6% said it raised it and 2.5% said it lowered it.
Most firms changed little to adopt AI. 64.3% of AI-using firms made no changes; 15.4% built new workflows, 15.0% trained staff and 3.8% used vendors or consultants to set it up.
Large firms put more into the change. 31.4% of AI users with 250 or more employees trained staff, against 13.4% of those with 1 to 4.
What holds owners back, and do plans turn into use?
Among firms with no plans to use AI, 61.6% said it does not apply to their business. The next reasons were not knowing what AI can do (22.0%) and privacy or security concerns (20.7%).
Cost was a minor reason: 6.9% said AI is too expensive. 3.2% said earlier use had disappointed them.
Plans have been a good guide to what happens next. Across 11 pairs of periods under the current question, the share expecting to use AI came within 0.9 points, on average, of actual use six months later.
Recent plans have been slightly too cautious. For plans made from February 2026 on, actual use six months later came in 1.3 to 1.6 points higher than expected.
Recommendations
- Compare yourself with your industry. Find your industry and size in Exhibits 2 and 4. In professional services, about four in ten firms with 1 to 4 employees use AI; in restaurants and hotels, fewer than one in ten.
- Start with the writing your team already does. Writing and editing is the most common use, at 85.4% of firms whose staff use generative AI. Pick one recurring document, such as quotes, follow-up emails or job reports, and set a house template and a review step.
- Look hard at the back office. Finance, HR and administration are where small firms trail large ones most. List the repetitive steps in billing, collections, bill entry and month-end reports, and pick the one with the clearest rules.
- Hand over one task at a time. Firms using AI mostly use it to support staff, and 95.7% saw no change in headcount. Choose one task, write down what a correct result looks like, and check the output for a few weeks before widening it.
- Settle the data question first. Privacy and security concerns hold back 20.7% of firms with no AI plans. Decide which customer and financial data may go into which tools, use business accounts with admin controls, and tell your team in writing.
- Write down the new way of working. Most AI users (64.3%) changed nothing when they started, while large firms trained staff more than twice as often. A one-page process note and a short team session are the small-firm version.
Method
We downloaded the Census Bureau’s published Business Trends and Outlook Survey (BTOS) estimates and the 2022 Statistics of US Businesses on November 8, 2026, and computed every figure in this paper with one script.
- Data: BTOS estimates for every biweekly period from September 11, 2023 to November 1, 2026, national and by industry, firm size, and industry by size [1]. About 200,000 businesses from a 1.2 million sample are asked each period; 24,000 to 25,000 responded in each of the last six comparable periods.
- Two series: the original AI question (“in producing goods or services”) ran from September 2023 to October 2025. A broader question (“in any of its business functions”) began on November 17, 2025 and reads higher, so we compare trends only within one series [4].
- Latest period: from October 19, 2026 the questionnaire calls the technology “Super Intelligence (formerly Artificial Intelligence or AI)”. “Do not know” answers rose from 9.6% to 13.9% that period, so our latest comparable reading is the survey collected October 5 to 18, 2026.
- Industry and size: to reduce noise, we average the six most recent comparable periods (collected July 27 to October 18, 2026), one full rotation of the sample. We compare them with the first six periods of the broader question (November 17, 2025 to February 8, 2026).
- Functions and tasks: the Bureau’s 2026 AI supplement pooled about 117,000 responses collected November 17, 2025 to February 8, 2026 [3]. Each question has its own base (all firms, AI users, planners or non-planners), which we state with each figure.
- Worker view: we combined the size-class rates with 2022 employment by firm size from the Statistics of US Businesses [2] to estimate the share of workers at firms using AI.
Limits
- Response rates are low: 12% to 13% nationally, and 8% to 9% for firms with 1 to 4 employees, in the last six comparable periods. The Bureau weights responses to represent all employer businesses, but non-response can still bias results.
- Answers are a self-reported yes or no. A firm that used AI once in two weeks counts the same as one that uses it all day.
- Farms and businesses without employees are outside the survey.
- Wording changes move the level of the series. The national reading went from 10.0% (September 2025, old wording) to 17.3% (November 2025, new wording), a jump the Bureau links to the new wording.
- Industry and size figures are simple averages of published biweekly estimates. Published standard errors are about 0.3 points for the national figure and 0.7 to 2.0 points for a single industry in one period.
- Some industry-by-size cells are suppressed in some periods. We show a cell only when at least four of the six periods are published.
- The worker estimate mixes 2022 employment counts with 2026 survey rates and treats every firm in a size class alike, so it is approximate.
Questions
What share of US small businesses use AI?
About one in four. 24.6% of responding employer firms used AI in the two weeks before the survey collected October 5 to 18, 2026. Firms with 1 to 4 employees averaged 23.3% over the latest six periods.
Which industries use AI the most?
Information (46.7%), professional, scientific and technical services (42.5%), finance and insurance (39.2%) and educational services (34.9%), averaged over July to October 2026. Accommodation and food services is lowest, at 9.4%.
Do small businesses use AI less than large ones?
Firms with 250 or more employees use AI more (40.7%). Firms with 1 to 4 employees (23.3%) are in line with firms of 5 to 49, and industry explains more of the difference than size.
What do small businesses use AI for?
Mostly sales and marketing, strategy, research and IT. Where staff use generative AI, 85.4% of firms use it for writing and editing, and about half to search for information.
Is AI causing layoffs at small businesses?
Rarely, so far. Among firms using AI, 95.7% reported no change in total employment from AI over six months, 2.3% an increase and 2.0% a decrease.
How reliable is the Census Bureau’s AI data?
It is a large, regular federal survey, with 24,000 to 25,000 responses every two weeks. Response rates are low (12% to 13%), answers are self-reported, and the 2025 wording change means old and new readings should be compared with care.
Citations
- [1] US Census Bureau, Business Trends and Outlook Survey (BTOS), data downloads: National, Sector, Employment Size Class, Sector by Employment Size Class and URR files (periods 202524 to 202622), and AI Core Questions (Original) (periods 202319 to 202520). Downloaded November 8, 2026. https://www.census.gov/hfp/btos/data_downloads
- [2] US Census Bureau, Statistics of US Businesses (SUSB), 2022 annual data by enterprise employment size, us_state_naics_detailedsizes_2022.txt. Downloaded November 8, 2026. https://www2.census.gov/programs-surveys/susb/tables/2022/
- [3] US Census Bureau, BTOS AI Supplement Table 2026 and AI Supplement 2026 response rates, responses collected November 17, 2025 to February 8, 2026. Downloaded November 8, 2026. https://www.census.gov/hfp/btos/data_downloads
- [4] US Census Bureau, “BTOS AI Core Question Updates” (December 3, 2025). https://www.census.gov/hfp/btos/downloads/AI%20Question%20Wording%20Updates.pdf
Walk us through one problem.
We’ll tell you what we’d fix first.
Walk us through one problemFree. 30 minutes. A one-page note within two business days.