Research/AI + Human Workforce

Business AI Use by Firm Size (2026): Census Survey Findings

Published 7 min read3 sources citedVerified 2026-11-04

Key Takeaways

  • In the Census AI supplement, 18.9% of firms with at least 10 employees reported current AI use, versus 35.1% of firms with at least 1,000 employees.
  • The published size-threshold rows overlap and use weighted firms as their denominators; they are not mutually exclusive employee bands.
  • The revised question asks about AI use in any business function during the prior two weeks, which breaks comparability with the older production-focused wording.
  • Reported use or expected use does not establish a reduction in staff hours, headcount, or costs.

Business AI use by firm size in 2026 rises as the size threshold rises, according to the U.S. Census Bureau's Business Trends and Outlook Survey (BTOS) AI supplement working paper. Among firms with at least 10 employees, 18.9% reported using AI in a business function in the prior two weeks. Among firms with at least 1,000 employees, the rate was 35.1%. Those are survey estimates for two overlapping groups of firms, not a forecast of what any one company should do.

The useful question for an operator is narrower than “Are we behind?” It is whether a particular workflow has enough volume, clear inputs, and a review process to justify an AI trial. Census measures reported use. It does not supply a target number of employees or a return on investment for your business.

Census AI use rates by firm size

Table C.3 of the Census working paper reports these firm-weighted estimates from the second BTOS AI supplement. The supplement was collected across six two-week panels from November 17, 2025, through February 8, 2026. “Current” refers to use in the prior two weeks; “future” is the business's expectation for the next six months.

Firms with this many employees or more Current AI use Expected AI use in next six months
10+ 18.9% 24.6%
50+ 24.3% 32.2%
100+ 27.3% 36.1%
200+ 30.6% 40.0%
500+ 32.4% 41.0%
1,000+ 35.1% 44.9%

Denominator: each percentage is the weighted share of surveyed businesses meeting that row's minimum employee count that answered yes to the relevant question. The rows are nested: the 1,000+ firms also appear in the 500+, 200+, 100+, 50+, and 10+ groups. Do not add the percentages or treat a row as the rate for an exclusive size band. The published table supplies percentages and average firm size, but no unweighted respondent count for each threshold. We therefore do not convert these estimates into counts of businesses.

The working paper's separate Figure 3 uses exclusive employee bands. Its text reports about 18% current use among firms with fewer than 20 employees and 31% among firms with at least 250 employees during the supplement period. These rounded figure values answer a different size-group question from Table C.3 and should not be inserted into the threshold table.

What changed in the question

The BTOS questionnaire with the AI supplement asks whether the business used AI “in any of its business functions” during the specified two-week period. The response choices are yes, no, and do not know. A separate question asks whether the business thinks it will use AI in any business function during the next six months. These are business-level responses, not a count of employees using a tool or a measure of how many tasks were automated.

The Census Bureau changed the core question on November 17, 2025. The previous wording asked about AI use “in producing goods or services.” The Census researchers explain that firms using AI for marketing, accounting, or project management might not have counted that use as production. The broader wording produced a level shift in reported use across size groups. A chart that connects rates from before and after the change as one uninterrupted trend would be misleading.

The supplement also asks about AI use across business functions and worker tasks. Those items have different reference periods and denominators. This article's table uses the two core business-level questions only.

How to read the size gap

The gap between the 10+ and 1,000+ current-use rows is 16.2 percentage points: 35.1 minus 18.9. That calculation compares overlapping threshold groups, so it is a descriptive contrast rather than a controlled estimate of the effect of being larger. Firm size, industry mix, work type, budgets, and respondent knowledge can all differ.

Expected use is also higher than current use in every displayed row. For the 100+ group, the difference is 8.8 percentage points: 36.1 minus 27.3. It is a gap between answers to two different questions at the same survey time, not observed growth six months later. Some plans may not be implemented.

The Census Bureau's May 2026 summary supplies a later, dated cross-check. For the collection period ending May 3, 2026, it reported 32% current use among firms with 100 to 249 employees and 37% among firms with at least 250 employees. It also reported a 19.8% national rate at that date. Those May figures are from a later BTOS collection and different size groups, so they should not be mixed into the supplement's Table C.3 as though all rows describe one date.

What the figures mean for a staffing decision

A small business can use this evidence to set a realistic discovery question, not a hiring target. For example, a 12-person service company could log the weekly volume of appointment requests, classify which requests follow a stable rule, and test a tool on a limited set. It would measure corrections, response time, customer outcomes, and the minutes staff spend reviewing outputs. If review takes as long as the original work, the trial has not freed capacity. None of those outcomes can be inferred from the BTOS adoption percentage.

A 300-person firm may have more places to test AI, but the same measurement rule applies. It can separate a draft-generation task from a decision that requires a person, assign an owner for quality checks, and compare the full workflow before and after a trial. The Census size gradient gives context for how common adoption is. It does not prove a larger firm can remove a role, or that a smaller firm needs to copy its tooling.

For a choice between software assistance and delegated human support, see our AI versus human virtual assistant comparison. The law-firm AI adoption study examines a specific profession with different survey populations. Its law-firm size percentages should not be substituted for these all-sector Census estimates.

Scope, limitations, and calculation method

BTOS is a recurring U.S. business survey. The Census working paper says the second supplement covered six biweekly panels from November 17, 2025, to February 8, 2026, using the survey's 1.2 million-business sample across the cycle and adjusting for nonresponse. That sample size is not the number of firms answering each table row. The paper does not publish row-level unweighted counts in Table C.3. Estimates are firm-weighted, so each firm's response contributes to an estimate of the share of firms, rather than the share of employees at firms using AI.

The survey asks a respondent to report business use; it does not audit software logs. “Any business function” includes activities with very different intensity, and a yes response does not tell us how often a tool was used. Future use is an expectation. The size comparison is descriptive and does not isolate cause. These constraints matter when converting a national survey result into a local operating decision.

We reproduced Table C.3's displayed firm-weighted percentages without recalculating the Census estimates. Our only arithmetic is subtraction of the displayed percentages: 35.1 - 18.9 = 16.2 percentage points, and 36.1 - 27.3 = 8.8 percentage points. Rounding in the source may affect comparisons at finer precision.

Sources

Checked November 4, 2026.

Tags

business AI use by firm size 2026Census BTOS AI supplementAI adoption by company size

Ready to put this into practice?

Book a free 15-min match call

Tell us what role you're filling. We'll match you with a pre-vetted virtual assistant - or tell you honestly if we're not the right fit.

Book a free call →

Related Research

Need Help Applying This to Your Business?

Book a free 15-minute match call. We'll recommend the right virtual assistant for your specific situation - no commitment required.

Book a 15-Min Match Call