Summary
AI can help small businesses and workers do more with less, but adoption is uneven. The smallest firms are falling behind. I think good policy should make adoption easier through community colleges and apprenticeships, set clear guardrails that small firms can afford to follow, and use AI openly and carefully to make public services faster and cheaper.
How many small businesses are actually using AI?
It depends on how you ask the question. The Census Bureau’s Business Trends and Outlook Survey found that between 17% and 20% of U.S. businesses used AI from December 2025 to May 2026. Between 20% and 23% expected to use it within six months (Census Bureau, 2026). The gap by size is large. About 37% of firms with 250 or more employees reported using AI. Fewer than 20% of firms with four or fewer employees did, and use did not change significantly among firms with fewer than 20 employees.
The Federal Reserve’s Small Business Credit Survey asked a broader question: whether the business or its employees currently use AI at all. By that measure, 46% of employer firms said yes and another 15% planned to start within a year (Federal Reserve Banks, 2026). Most of that use is early. About half of AI users are still experimenting, 44% have partly built it into their work, and only 7% have fully integrated it. These two surveys use different samples and questions, but together they point to one conclusion. Many owners are trying AI, and relatively few have built it into how the business runs.
What are small firms getting out of it?
According to the same Federal Reserve survey, small businesses most often use AI for writing and marketing (83% of users), individual productivity (61%) and planning or analysis (51%). Seventy-one percent of users said AI increased productivity, 39% said it improved the quality of their goods or services, and 31% reported higher sales. Most said it had not changed their labor costs. The top problems were accuracy (46%) and adapting tools to their needs (43%). Among firms planning to adopt AI, the biggest obstacles were finding the right tools (54%) and the time needed to set them up and train staff (37%).
My view is that the main obstacle is know-how, not access to AI tools. A large company can hire a team to test and implement AI. A ten-person business can’t. That is where public policy can help without picking winners.
How should we prepare workers?
The best training institutions for this already exist. There are 1,015 community colleges in the United States, and they enrolled about 10.6 million students in fall 2024, including 4.1 million in noncredit programs (American Association of Community Colleges, 2026). Noncredit, short-format courses are well suited to teaching practical AI skills to working adults and small business owners.
Registered apprenticeships are the other proven model. The Department of Labor reports more than 800,000 apprentices a year, an average starting salary of $86,000 for people who complete a program, and 93% of completers staying employed (Apprenticeship.gov). I would expand apprenticeships in fields such as IT support, data work, cybersecurity and AI-assisted trades. I would also let groups of small employers sponsor apprentices together so no single firm has to carry all the paperwork.
What guardrails make sense without crushing small firms?
Guardrails should be based on risk and matched to how the AI is actually used. The National Institute of Standards and Technology released its AI Risk Management Framework on January 26, 2023, as a voluntary, consensus-based guide, and added a companion profile for generative AI on July 26, 2024 (NIST). A framework like that gives businesses a shared vocabulary without requiring every company to hire compliance staff.
In practice, I think the strictest requirements belong where AI helps make high-stakes decisions about people, such as hiring, lending, housing, health care and public benefits. Businesses using AI for those decisions should test for errors and bias, keep records and give people a way to reach a human. A bakery using AI to write menu descriptions should not face the same paperwork. Rules should also be consistent from state to state wherever possible, because a small firm cannot afford to follow 50 different versions.
Can AI make government work better?
Yes, if it is used carefully. Many residents deal with government through forms, permits, benefit applications and call centers. AI can help staff sort applications, draft routine replies, translate information and find errors before they turn into backlogs. It should not make final decisions about someone’s benefits or liberty without a person reviewing them. Agencies should also publish which AI tools they use, what those tools are for and how they are tested. If an AI tool makes a service faster and more accurate, the public should be able to see it. If it does not, it should be dropped.
Is transparency good for business too?
I think so. Customers are more likely to trust products when they know when they are dealing with AI and how their data is used. Clear disclosure rules, especially for AI-generated content and automated decisions, also protect honest businesses from competitors who cut corners. Transparency works best when requirements are simple, standardized and proportionate to the stakes.
What I’d prioritize
- Fund short, noncredit AI courses at community colleges for small business owners and their staff.
- Expand registered apprenticeships in technology roles and let small employers sponsor apprentices together.
- Make the strictest AI rules apply to high-stakes decisions, and base those rules on the NIST framework.
- Push for consistent AI rules across states so small firms don’t have to follow 50 different versions.
- Require public agencies to disclose the AI tools they use and to keep human review for consequential decisions.
Frequently asked questions
What share of small businesses use AI?
Estimates vary with the question asked. Census data show 17% to 20% of all businesses using AI in late 2025 and early 2026. The Federal Reserve’s 2025 survey found that 46% of employer firms or their employees use AI in some form.
Will AI eliminate small business jobs?
The evidence so far doesn’t show large effects. Most small-firm AI users in the Federal Reserve survey reported no change in labor costs, and 71% reported higher productivity.
What is the NIST AI Risk Management Framework?
It is a voluntary federal framework, released in January 2023, that helps organizations identify and manage AI risks. A generative AI profile was added in July 2024.
Sources
- U.S. Census Bureau, “AI Use at U.S. Businesses” (May 2026): census.gov
- Federal Reserve Banks, “2026 Report on Employer Firms: Findings from the 2025 Small Business Credit Survey” (2026): fedsmallbusiness.org
- American Association of Community Colleges, “Fast Facts 2026” (2026): aacc.nche.edu
- U.S. Department of Labor, Apprenticeship.gov: apprenticeship.gov
- National Institute of Standards and Technology, “AI Risk Management Framework”: nist.gov