Which service industries are using AI, by the Census numbers

About one in four U.S. employer businesses now uses AI somewhere in its operations. The Census Bureau's Business Trends and Outlook Survey (BTOS) asks a national sample of businesses every two weeks whether they used AI in any business function in the past two weeks. In the wave collected October 5 to 18, 2026, 24.6% said yes and 28.6% expected to use it within six months.

Adoption is concentrated in industries built on information and paperwork. Nearly half of information firms (46.4%) and professional, scientific and technical firms (43.8%) use AI, along with 40.5% of finance and insurance firms. Industries built on hands-on, in-person service use it far less: 14.6% of repair and personal services firms and 10.7% of accommodation and food service firms.

Finance and insurance, health care and professional services gained the most since the Census began its current AI question in November 2025. Finance and insurance rose 10.6 points, from 29.9% to 40.5%. Health care rose 9.5 points, from 18.1% to 27.6%, and professional services 9.1 points, from 34.7% to 43.8%. Real estate (up 8.0 points) and administrative and support services (up 7.4) also grew faster than the economy-wide average of 7.3 points.

In practice, that means an insurance agency, a CPA firm or a physical therapy group is now more likely than not to have peers using AI for tasks such as drafting, summarizing, data analysis or answering routine customer questions. Food service and lodging and educational services barely moved; their changes are within the survey's margin of error.

Expectations point to more growth. In every service sector, the share expecting to use AI within six months is higher than current use. In finance and insurance, 47.4% expect to use it; in professional services, 48.8%; in health care, 32.9%.

Yes, sharply at the top. In the October 2026 wave, 43.6% of firms with 250 or more employees used AI, against 24.3% of firms with 1 to 4 employees and 22.6% of firms with 5 to 9. The gap between the smallest and largest firms is roughly 20 points.

The middle is closer together than many expect. Firms with 10 to 19 employees (24.6%) and 20 to 49 employees (26.0%) use AI at about the same rate as the smallest firms. The step up comes at 50 employees and above. A Census Bureau story in May 2026 reported the same pattern, noting that use rose among firms with 20 or more employees between December 2025 and May 2026 while firms with fewer than 20 showed no significant change over that window. By October, the smallest firms had also moved up, from 17.0% in November 2025 to 24.3%.

South Carolina is in line with the nation. In the October 2026 wave, 27.7% of South Carolina employer businesses reported using AI and 34.6% expected to within six months. State estimates come from smaller samples and swing between waves (South Carolina's ranged from 20.4% to 27.7% across the last four), so a four-wave average is a better guide: 23.8% for South Carolina against 23.9% nationally.

Neighboring states fall in a similar band. In the latest wave, North Carolina reported 29.6%, Florida 26.8%, Georgia 25.4%, Tennessee 24.9% and Alabama 22.1%. The Census does not publish state estimates by sector, so we cannot say whether South Carolina's health care or legal firms differ from national peers.

If you run a law firm, accounting practice, insurance agency or consulting business, AI use among your peers is now common: 40% to 44% of firms in those sectors use it. If you run a restaurant, salon, auto shop or event venue, most peers do not use it yet, so careful adoption can still set you apart.

The survey measures whether a business uses AI at all, not how well. The useful question for a service business is narrower: which repeated tasks take staff time without needing their judgment? Common examples include drafting first replies to inquiries, summarizing intake notes, sorting incoming requests by urgency, and answering after-hours questions about hours, pricing ranges or availability. Each should have a clear owner, a review step and a way to hand off to a person.

Monvell's Automation team builds these workflows with that structure, and our Consulting work starts by identifying which tasks are worth automating at all.

• List the five tasks your team repeats most often each week and note how long each takes.
• Mark which ones need professional judgment and which follow a predictable pattern.
• Pilot AI on one predictable task with a human review step before anything reaches a customer.
• Check your industry's rules first: health care, legal and financial firms have privacy and confidentiality duties that apply to any tool handling client data.
• Measure time saved and error rates before expanding to another task.

Dataset: U.S. Census Bureau Business Trends and Outlook Survey (BTOS), national, sector, employment size class and state download files, accessed November 1, 2026. BTOS surveys about 1.2 million employer businesses (excluding farms) in six rotating panels, with collection every two weeks.

Question: 'In the last two weeks, did this business use Artificial Intelligence (AI) in any of its business functions?' and the matching six-month expectation question. The Census changed the wording on November 17, 2025, from 'in producing goods or services' to 'in any of its business functions', and started a new time series because the change caused a level shift. We therefore compare the latest wave (collection October 5 to 18, 2026, reference period September 21 to October 4) with the first wave of the new series (collection November 17 to 30, 2025), about 11 months apart, rather than a full year.

Sectors: NAICS 51, 52, 53, 54, 56, 61, 62, 71, 72 and 81. We computed percentage point changes and flagged changes smaller than an approximate 90% margin of error (1.645 times the square root of the sum of the two published squared standard errors). This simple test ignores correlation between waves.

Limits: BTOS is a voluntary survey with weighted estimates; sector estimates in smaller industries have standard errors of 1 to 3 points, and state estimates of 2 to 5 points. Data for October 6 to November 16, 2025 were not collected because of the federal funding lapse. 'Use' is self-reported and covers anything from occasional drafting help to core operations.