Ask five sources how many small businesses use AI and the answers range from under 20% to over 75%. That spread is not a data error. The U.S. Census Bureau measures a narrow slice, AI used directly in producing goods or services, while owner-survey outfits like Goldman Sachs count anyone who has run a draft through ChatGPT. This piece works through four dated, named surveys, plus California’s own economic data, to answer plainly: what adoption looks like now, who’s ahead, what’s stopping the rest, and what closes the gap.
What is the current AI adoption rate among small businesses?
The honest answer depends on how “using AI” gets defined, and that single choice explains most of the gap between headlines. The U.S. Census Bureau’s Business Trends and Outlook Survey (BTOS) added AI-use questions between November 17, 2025 and February 8, 2026, and its biweekly data put overall U.S. business AI use at 17% to 20% from December 2025 through May 2026. That figure uses a strict definition: AI used directly in producing goods or services, not just present somewhere in the office.
Owner-survey outfits count differently and land much higher. The Goldman Sachs 10,000 Small Businesses Voices survey, fielded by Babson College and David Binder Research from January 27 to February 4, 2026 among 1,256 small business owners, found 76% currently using AI in some form. The U.S. Chamber of Commerce’s Empowering Small Business report put generative AI use at 58%, up from 40% in 2024 and more than double the 2023 rate.
Both numbers are accurate; they measure different things. BTOS asks whether AI touches core production. Goldman Sachs and the Chamber ask owners whether AI touches the business at all, which counts a bookkeeper drafting emails in ChatGPT the same as a firm running AI in its actual service delivery.
Read together, the real picture is this: most owners have tried AI somewhere in the business, a much smaller share have it built into how work actually gets done, and very few call it finished. Only 14% of Goldman Sachs’s own respondents describe AI as fully embedded in core operations, even inside a survey where 76% say they use it.
Which businesses are ahead, and which are behind?
Firm size splits the data more sharply than industry does. The U.S. Chamber of Commerce Foundation’s Main Street AI Monitor, an Ipsos survey of 1,070 small-business workers fielded May 8 to 11, 2026 and published June 17, 2026, found 43% of workers at businesses with 2 to 9 employees using AI at work, against 59% at businesses with 100 to 249 employees. That 16-point gap tracks headcount, not sector.
Census BTOS tells the same size story from the employer side. AI use kept climbing among firms with 20 or more employees between December 2025 and May 2026, while firms with four or fewer employees stayed flat over the same period, with fewer than one in five reporting any AI use at all.
Industry still matters, just less. Census’s most recent data puts the Information sector at 39.7% current AI use and Finance and Insurance at 33.9%, both well ahead of Retail Trade at roughly 14%. That order tracks which sectors already run on structured data and repeatable digital workflows, rather than physical inventory and in-person transactions.
A California-specific adoption percentage doesn’t exist yet in the public survey record, but the county data already compiled for this program’s training playbook offers circumstantial evidence, not a substitute for a measured rate. Bureau of Labor Statistics Quarterly Census of Employment and Wages (QCEW) 2024 annual data shows Professional, Scientific, and Technical Services as the second-largest sector by establishment count in both Santa Clara County (10,012 establishments) and San Francisco County (8,888 establishments), with average weekly wages of $5,598 and $4,901, among the highest of any sector in either county’s data. That’s knowledge-intensive, digital-first work concentrated in two of the state’s largest metro economies, which points the same direction as the national industry pattern above without claiming to measure it directly.
What’s actually holding small businesses back?
Cost used to top every list of objections. It is no longer the leading barrier in 2026 survey data; clarity and skill are. The Small Business & Entrepreneurship Council’s March 2026 Small Business Technology Use Survey found limited knowledge or training was the top barrier to further AI adoption, ahead of both cost and data-privacy concerns, even though the survey’s own respondents report a median annual AI spend of just $2,200.
The Chamber Foundation’s Main Street AI Monitor found the same pattern from a worker’s-eye view. 41% said it is unclear how AI applies to their specific job, and a similar share cited a skills gap outright as the reason they haven’t gone further.
Formal training remains rare against that need. Only about one in ten workers surveyed by the Chamber Foundation said they had ever received any. Goldman Sachs’s data echoes the gap from the owner side: 45% of AI users cite a lack of technical expertise as a real obstacle, and 47% say it is hard to choose the right tool among the options on the market.
The starkest version of this pattern sits at the smallest end of the market. Earlier Chamber research found 82% of firms with fewer than five employees believe AI “isn’t applicable” to their business, which the researchers themselves called an education gap rather than a real limitation. Time is the barrier hiding behind both numbers: a solo owner or a five-person team has no spare hours to self-teach a new tool category between client work, which is exactly the gap a structured session is built to close rather than a cost problem money alone would fix.
How training changes the calculus
Training is what turns “someone on the team tried ChatGPT once” into the thing that reliably predicts results: embedded, repeatable use. Goldman Sachs’s own numbers show why the gap matters. 76% of small business owners use AI, but only 14% call it fully embedded in core operations, and that gap is a training and process problem, not a technology one.
The payoff for closing it shows up in the same surveys. SBE Council’s 2026 respondents who invested in AI training report a median of five hours saved per week personally and 11.5 employee hours saved per week organization-wide. Two-thirds report AI-attributable revenue gains, and roughly one in five report gains above 10%.
Owners themselves rank training as what they want most to make the next step work. SBE Council’s survey put practical, hands-on training ahead of cheaper tools or clearer regulation as the top support need small businesses want.
For a California business, that training carries a funding mechanism most owners have never used. The state’s Employment Training Panel (ETP) Small Business Program reimburses qualifying training hours at $28 per trainee hour for California-domiciled firms with 100 or fewer California employees, funded by a payroll tax most employers already pay into. The California AI training playbook walks through eligibility and the application mechanics in full.
What a first step looks like
The right first step already exists for most California owners; it is underused. Free public sessions, run by regional SBDC offices, SCORE chapters, chambers of commerce, and vendor-backed events like the Bay Area Council’s AI for All initiative and Meta’s Small Business Accelerator, remain a reasonable and low-risk way to see what’s possible before spending anything. We recommend treating a free session as the first step, not the destination.
What those sessions can’t do is change how one specific team does its specific work. Leah Toeniskoetter, CEO of the San Jose Chamber of Commerce, described the confusion many owners bring to these events at a Meta-hosted small-business AI discussion there, as NBC Bay Area reported in September 2025: “I think small businesses start with confusion. What is it? It means so many things.” That reaction lines up with the 41% of small-business workers nationally who say it’s unclear how AI applies to their specific job.
A structured workshop exists to answer that specific-job question instead of the general one. The corporate AI training guide for Los Angeles breaks down what that looks like in one of California’s largest small-business markets: group size, delivery format, and the industries it’s built around. The mechanics are consistent statewide: a facilitated program built from the firm’s own documents, priced in the $2,000 to $15,000 market range for a workshop, with ETP reimbursement able to offset a real share of that cost.
Frequently asked questions
How many small businesses actually use AI in 2026?
It depends on the definition. The Census Bureau’s BTOS, limited to AI used directly in producing goods or services, measured 17% to 20% of U.S. businesses from December 2025 through May 2026. Owner-survey outfits with a looser definition report much higher numbers: 76% in Goldman Sachs’s 10,000 Small Businesses Voices survey (January to February 2026) and 58% in the U.S. Chamber of Commerce’s Empowering Small Business report.
Why do AI adoption statistics vary so much between sources?
Mainly because “using AI” isn’t a single measured behavior. Census BTOS asks whether AI is used directly in producing a firm’s goods or services, a narrow bar that excludes a lot of everyday tool use. Owner-survey methods (Goldman Sachs, the U.S. Chamber, SBE Council) ask whether AI touches the business in any way, which counts a ChatGPT-drafted email the same as AI inside core service delivery. Neither method is wrong; they’re answering different questions.
Which industries have the highest AI adoption rates?
The Information sector leads Census’s most recent industry breakdown at 39.7% current use, followed by Finance and Insurance at 33.9%. Retail Trade sits far behind at roughly 14%, tracking which sectors already run on structured, digital-first workflows rather than physical inventory and in-person transactions. California’s own BLS QCEW county data shows similar knowledge-intensive work (Professional, Scientific, and Technical Services) concentrated in Santa Clara and San Francisco counties, though that’s circumstantial context, not a measured California rate.
Does firm size matter more than industry for AI adoption?
Yes, based on 2026 survey data. The Chamber Foundation’s Main Street AI Monitor found 43% of workers at businesses with 2 to 9 employees using AI, against 59% at businesses with 100 to 249 employees, a 16-point gap that tracks headcount rather than sector. Census BTOS shows the same pattern from the employer side: AI use kept rising among firms with 20 or more employees between December 2025 and May 2026, while firms with four or fewer employees stayed flat over that same period.
Is there California-specific data on small business AI adoption?
Not yet as a single named statistic. Census BTOS collects data nationally, by industry, and by firm size, but the published AI Use Supplement results don’t report a distinct California percentage. What exists instead is context: BLS QCEW county data shows Santa Clara and San Francisco counties with an outsized share of Professional, Scientific, and Technical Services establishments, the kind of knowledge-intensive work tied to faster adoption nationally, which is a reasonable basis for an inference, not a substitute for a measured state figure.
What’s the biggest reason small businesses haven’t adopted AI yet?
Not knowing where to start or how it applies, more than cost. SBE Council’s March 2026 survey found limited knowledge or training was the top barrier to further adoption, ahead of cost and data-privacy concerns, despite a median annual AI spend of only $2,200 among respondents. The Chamber Foundation’s Main Street AI Monitor found the same thing from workers directly: 41% said it’s unclear how AI applies to their specific job, and a similar share cited a skills gap.
Does AI training actually pay for itself for a small business?
Survey data says yes, more often than not. SBE Council’s 2026 respondents who invested in AI training report a median of five personal hours saved per week and 11.5 employee hours saved per week organization-wide, with two-thirds reporting AI-attributable revenue gains. For California businesses, the ETP Small Business Program can reimburse qualifying training hours at $28 per trainee hour, which changes the math further before a firm even counts the hours saved.
Key takeaways
- Adoption numbers that look contradictory (under 20% versus over 75%) usually aren’t; they’re measuring different things. Census BTOS uses a strict production-use definition; owner surveys count any AI use at all.
- Firm size predicts adoption more reliably than industry does. The gap between a 2-to-9-employee shop and a 100-to-249-employee company (43% versus 59% AI use, per the Chamber Foundation’s Main Street AI Monitor) is wider than the gap between most industries.
- Cost has receded as the top objection. Limited knowledge, unclear application, and a lack of formal training now rank ahead of price in 2026 survey data from SBE Council and the Chamber Foundation.
- The gap that matters most is between “using AI” and “AI embedded in operations.” Goldman Sachs’s own data shows 76% adoption but only 14% full embedding, and that gap is what structured, role-specific training is built to close.
- California businesses have a funding mechanism most competitors in other states don’t: ETP Small Business Program reimbursement at $28 per trainee hour for qualifying training.
Businesses that want a straight read on where their own team stands, and what a structured first step would look like for their specific roles, can book a free AI-readiness call or start with the overview on the Team Training page.
Dirk Jan van Veen, PhD
