Two surveys published this year describe the same room from opposite ends. The U.S. Chamber of Commerce reports that two thirds of American small businesses now use AI. SAS and IDC report that only 9% of small and midsize businesses have fully embedded it into strategy, operations and decision-making.
Both can be true at once, and that is the useful part. Using AI and running on AI are different states, and the distance between 66% and 9% is where most small business AI money is currently being spent without a return.
Start with the numbers as their publishers state them, because they are not measuring the same thing.
The U.S. Chamber of Commerce 2026 report says 66% of U.S. small businesses use AI, up 8 points from 2025 and nearly three times the 2023 share of 23%. It also reports that 41% now use six or more technology platforms, up 9 points from 2025. That is a count of adoption, not of results.
The SAS and IDC research, published in May 2026 and based on a survey of more than 1,600 SMB leaders across 28 countries, found that nearly 70% of small and midsize businesses remain in the experimental or opportunistic stages of AI maturity. Its companion report page is blunter: 37% are still experimenting in isolated, unconnected ways, and only 9% have fully embedded AI into strategy, operations and decision-making.
Read the middle row carefully. Nearly 70% is the combination of the first stage and the stage above it, not a third bucket alongside them. A business that has an AI tool in three departments and cannot say what it changed is in that 70%, even if its owners would describe the company as "using AI".
The Chamber report also measured what small businesses say AI is doing to employment, and the answer is not the one the headlines assume. 47% say AI is creating jobs today, while 6% say it is enabling headcount reductions. 95% of small businesses using AI said they are working to upskill employees.
It also reports that small businesses using AI are at least 7 points more likely to report sales, profit and workforce growth than those that are not. That sentence is a correlation from a survey, not a causal claim, and it should be read as one. Businesses already growing are more likely to invest in tools. The Chamber does not claim AI produced the growth.
An isolated tool produces an isolated result. Nobody owns it, so nothing improves when it is wrong. Nothing is measured, so nobody can tell whether it helped. When the person who set it up leaves, it either breaks quietly or keeps running unexamined.
That is the readiness gap SAS describes, and it is an operations problem rather than a model problem. The tool is rarely the weak link. The missing pieces are a named owner, a written process, data the agent can actually reach, and a measurement that outlives the person who set it up. Those are the same gaps covered from the enterprise side in why SMB AI agent adoption lags enterprise.
Pick one function that already runs on a repeatable process, with a number attached and a person who cares about that number. Put an agent on the whole function rather than a step of it. Define what it may do alone and what stops for a person, following the pattern in bounded permissions and human approval. Write the process down before building it, so the spec exists independently of the tool, which is the discipline in spec-driven development with AI agents.
Then run it for a month and compare the number. That is the difference between the 37% and the 9%. The businesses in the embedded group are not using cleverer models; they have a function that runs, an owner, and a review that happens whether or not anyone feels like it. The practical roles are described in AI agents for small business and back-office AI agents going mainstream.
Survey data is self-reported. A respondent who answers "yes" to using AI may mean a chat window, and one who answers "fully embedded" is grading their own homework. The two studies also sample differently, so the 66% and the 9% are not two ends of one measurement. Do not put them in a single chart as a trend line.
What they agree on is the shape: widespread use, thin integration. If your own business recognises itself in the 37%, the surveys have done their job and the rest is execution, which is an operational question before it is a technical one.
No. Embedding means AI runs defined parts of operations with an accountable owner and a review cadence. A tool in three departments with no shared data and no measurement sits in the large early-stage group.
Yes. The 9% comes from the SAS and IDC research on small and midsize businesses, based on a survey of more than 1,600 SMB leaders across 28 countries. The 66% comes from a separate U.S. Chamber report on U.S. small businesses.
No. They report what respondents said. 47% said AI is creating jobs and 6% said it is enabling headcount reductions. That is opinion data from a survey, and the growth correlation is not evidence of cause.
Name one function, one number and one owner, then put the agent on the whole function rather than a step. Measurement and accountability are what separate the embedded group from the rest, not model choice.
Maxpertise builds and runs the AI team that operates your business: leads in, clients managed, invoices out. The managed lane exists precisely for the businesses the SAS research describes, with no technical person on staff and no appetite for running infrastructure. The embedded lane puts AI engineers inside an existing engineering team when the client directs the work.
Both lanes run on the same two human gates. A person approves the plan before anything is built, and a person reviews the result before it ships. The roles that make up the team are on the AI team page, the method and the gates are at how it works, and the engagement options and what each includes are on pricing.
If you recognise your business in the 37% and want a function that runs rather than another tool, start with a written intake at contact. Describe the process and the number you want to move; that is enough to scope it.
[1] U.S. Chamber of Commerce, Empowering Small Business: The Impact of Technology on U.S. Small Business, 2026. The 66% adoption figure, the 8 point rise from 2025, the 23% 2023 comparison, the 47% and 6% jobs findings, the upskilling figure, and the growth correlation.
[2] SAS, Majority of SMBs remain in early stages of AI maturity, study says, May 14, 2026. The near-70% early-stage finding and the survey base of more than 1,600 SMB leaders across 28 countries.
[3] SAS, AI readiness report for small and midsize businesses, 2026. The 37% experimenting in isolated ways and the 9% fully embedded figures.
Maxpertise is an AI-native engineering company. We embed native AI engineers inside your team, live in about 10 days. Please enable JavaScript to view the site, or email contact@maxpertise.net.