How AI is changing entrepreneurship: what the data shows so far
The short answer: AI is changing what one person can produce, not what makes a business succeed. Roughly one in five U.S. businesses now uses AI in some function, with large firms adopting at nearly twice the rate of the smallest. Among small employers using it, most report productivity gains, but very few have integrated it deeply, and most have seen no change in labor costs. The number of one-person businesses has been rising for decades, and AI extends what those businesses can do alone. What has not changed: survival rates, the difficulty of reaching customers, and the need for judgement about what to build and for whom.
Adoption: real, uneven, and led by larger firms
The most reliable measure of business AI use in the United States is the Census Bureau's Business Trends and Outlook Survey, a biweekly, nationally representative survey that asks whether a business used AI in any business function.
Between December 2025 and May 2026, overall AI use hovered between 17 and 20 percent of businesses, with 20 to 23 percent expecting to use it within six months. As of May 3, 2026, the national rate was 19.8 percent.
Size matters a great deal. Thirty-seven percent of firms with at least 250 employees reported using AI, and 32 percent of firms with 100 to 249 employees. Among firms with four or fewer employees, fewer than 20 percent did. Use rose among firms with at least 20 employees over that period and did not change significantly among smaller ones.
Sector matters too. Information businesses reported 39.7 percent use and finance and insurance 33.9 percent, against about 14 percent in retail trade.
The SBA's Office of Advocacy, drawing on the same survey for an earlier period, noted that the largest firms have outpaced the smallest in adoption but that the gap has narrowed.
Stanford's 2026 AI Index, which draws on surveys of organizations that skew toward larger companies, reports adoption of 88 percent among surveyed organizations, with generative AI used in at least one business function at 70 percent. The same report notes that deployment of AI agents, systems that act rather than draft, "was in the single digits across nearly all business functions."
Put together: nearly every large organization is using AI somewhere, a fifth of all U.S. businesses are, the smallest businesses lag, and almost nobody has autonomous AI doing real work yet.
What small businesses get from it
The Federal Reserve Banks' 2025 Small Business Credit Survey, with 6,525 responses from firms with 1 to 499 employees, gives the most detailed picture of small-firm use. Forty-six percent of firms used AI, and 15 percent more planned to.
The uses were concentrated: writing or marketing (83 percent of users), individual productivity (61 percent), and planning or analysis (51 percent). The reported results: 71 percent said productivity increased, 39 percent saw better quality of goods and services, and 31 percent reported higher sales.
Depth of use was low. About half of users were experimenting, 44 percent had partially integrated AI, and only 7 percent had fully integrated it. The main obstacles were accuracy (46 percent) and adapting tools to the business (43 percent).
One finding cuts against the loudest predictions: most firms using AI had not experienced changes to their labor costs as a result. The gains, where they exist, show up as time and output, not as smaller payrolls.
The trend AI is accelerating: the one-person business
The structural shift that predates AI is the rise of businesses with no employees. The Office of Advocacy counts 36.2 million small businesses, of which 82.3 percent, about 29.8 million, are nonemployer firms. That number has nearly doubled from 15.4 million in 1997. Small businesses as a whole employ 62.3 million people, 45.9 percent of the private workforce, and produce 43.5 percent of GDP.
AI fits this trend precisely. The tasks small firms report using it for, writing, productivity, planning, are the tasks a solo operator used to either do badly, do late, or hire out. A one-person business can now produce the proposals, marketing, documentation, and admin of a larger one. The ceiling on what one person can run has moved up.
What has not moved is the floor. New business formation remains high; the Census Bureau counted 497,046 seasonally adjusted business applications in December 2025 alone. Survival rates, measured over 1994 to 2022, show 67.7 percent of new employer establishments reaching two years, 49.2 percent reaching five, and 33.9 percent reaching ten. There is no evidence yet that AI has changed those odds, and no reason to expect it to, because the reasons businesses fail are not the reasons AI helps.
What is actually changing
Reading across the sources, four shifts are visible.
The cost of a first draft is approaching zero. Copy, plans, code, designs, and analyses that used to be the expensive part of starting something are now cheap. That removes a barrier to entry, which means more entrants and more noise. The advantage moves to whoever can tell a good draft from a bad one.
Capability is being bought instead of hired. The "adapting to the business" problem is the same problem consulting has always solved, and AI has made it the central skill. The Fed data on unchanged labor costs suggests firms are not replacing people; they are extending what existing people can do.
The gap between large and small is a gap in integration, not access. Everyone has the same tools. Large firms have the staff to embed them in processes. The 7 percent full-integration figure among small firms is the measurable size of that gap, and it is closable, because integration is mostly a matter of choosing one task and designing the check.
Trust is becoming the scarce asset. When anyone can produce polished output, polish stops signalling competence. Specific claims, cited sources, real examples, and a person who answers become the differentiators. Businesses that were already run that way gain; businesses that relied on looking bigger than they were lose.
What has not changed
Reaching customers is still the most commonly reported operational challenge among small employers. Rising costs are still the most common financial one. More than half of new establishments still close within five years. Judgement about what to sell, to whom, at what price, is still the founder's job, and an assistant that is designed to be agreeable is a poor substitute for a customer who is not.
AI is a change in the cost of production. Entrepreneurship was never mainly a production problem.
Frequently asked questions
Is AI creating more businesses? Business applications remain high, but they were high before generative AI arrived, and the data does not separate the causes. What can be said is that the cost of starting has fallen for anything that is mostly writing, planning, or software.
Is AI replacing small business employees? Not according to the Fed survey, where most AI-using firms reported no change in labor costs.
Which businesses benefit most? Those whose work involves a lot of repetitive text and decisions that a person can check quickly, and those with the discipline to integrate one task at a time. Sector data suggests information and financial businesses lead; retail lags.
Where should a small business start? With one repetitive, well-defined task and a clear way to check the output. The full-integration figure is low precisely because most start with everything at once.
Sources. U.S. Census Bureau, Large Firms With at Least 20 Employees Biggest AI Users, May 26, 2026, Business Trends and Outlook Survey. U.S. Census Bureau, Business Formation Statistics, December 2025 release. U.S. Small Business Administration, Office of Advocacy, Frequently Asked Questions About Small Business, February 2026. Stanford Institute for Human-Centered AI, The 2026 AI Index Report, Economy chapter. Federal Reserve Banks, Small Business Credit Survey, 2026 Report on Employer Firms. All accessed September 2026.