The AI Gap: 46% of Small Businesses Use AI. 31% Say It Raised Sales.

Key findings

• 46% of employer firms use AI. Another 15% plan to start within 12 months.

• 7% of AI users have fully integrated it into the business. 49% are still experimenting.

• 71% of AI users report increased productivity. 31% report increased sales.

• 83% of AI users use AI for writing or marketing. 29% use it for customer service.

• The top challenge named by AI users is accuracy, at 46%.

Source for all figures: Federal Reserve Banks, 2026 Report on Employer Firms (2025 SBCS). Full citations at the end of this post.

Nearly half of American small businesses now use AI somewhere in the operation. The Federal Reserve's latest Small Business Credit Survey puts adoption at 46% of employer firms, with another 15% planning to start within a year. Ask what the tools have delivered so far and the numbers get quieter. Among firms that use AI, 31% say sales went up. The group reporting a productivity gain is 71%, more than twice the size of the group reporting a sales gain.

Adoption is broad. Depth is thin.

Sixty-three percent of AI users told the Fed the tools are somewhat or very important to producing their core goods or services. The same owners describe a shallow footprint. Forty-nine percent of AI users say they are still experimenting — testing tools, running pilots. Forty-four percent have worked AI into some of their processes. Seven percent have embedded it across business functions.

So the typical small business in 2026 has AI in the building. It lives in a handful of workflows, used by a handful of people, while large parts of the operation run the way they ran before it arrived.

The work AI does is mostly desk work

Writing and marketing leads the list of uses at 83% of AI-using firms, followed by individual productivity at 61% and planning or analysis at 51%. Administrative functions come in at 38%. Customer service sits near the bottom at 29%, and process automation is lower still at 21%.

The pattern is easy to read. Owners bought AI to help with the work on their screens: drafts, summaries, schedules. The work that faces customers — the phones, the inbox at 9 p.m. — is the least automated part of the average firm.

That split shows up in the results owners report. Productivity rose for 71% of AI users. Quality of goods and services improved for 39%. Sales rose for 31%. And 77% said AI had not changed their labor costs at all. Taken together, the results describe a tool that makes existing work faster. Faster drafts are worth having. They do not answer a ringing phone, and they do not reach the customer who called twice and then booked a competitor.

Owners who use AI expect more, and the gap survives scrutiny

Owners who use AI expect a stronger year than owners who do not. Economists at the Federal Reserve Bank of New York measured this as a net score: the share expecting growth minus the share expecting decline. On revenue, AI users scored +48 against +21 for non-users. On employment, the scores were +33 against +15.

Some of that gap reflects the firms themselves. Younger owners tend to expect more, and so do firms with more than ten employees. After the researchers accounted for firm and owner characteristics and for each firm's recent track record, AI users kept a 14-point advantage in expected revenue growth and a 10-point advantage in expected employment growth.

The gap runs widest where conditions are hardest. Among firms describing their finances as fair or poor, the revenue-expectations gap between AI users and non-users reaches 24 points, compared with 15 points among firms in good shape. AI users who report trouble utilizing technology — websites, social media, e-commerce, security — post revenue expectations 16 points higher than AI users without that problem. The New York Fed's read on this is measured: AI may be helping firms work around specific constraints that were holding back revenue.

The authors also state the limit plainly. These figures are expectations gathered in the fall of 2025. Whether the firms actually hit those numbers is, in their words, a subject for future research.

What slows firms down

Ask current users what gets in the way and the list is practical. Accuracy comes first, named by 46%. Owners worry about bias and wrong information reaching customers. Close behind, at 43%, comes the work of finding tools and fitting them to the business. Firms planning to adopt next year put tool-finding first at 54%, and 37% name the time it takes to implement AI and train employees on it.

None of this reads as resistance to the technology. It reads as a capacity problem. A five-person shop has no IT department to hand the project to. The owner is the IT department, along with everything else.

Context from the same survey keeps the picture honest. Fifty-seven percent of firms named reaching customers and growing sales as an operational challenge in the prior year, the most common answer on the list. Twenty-nine percent named utilizing technology. The broad mood is cautious: expectations for revenue and employment growth across all firms sit at their lowest levels since the 2020 survey. Owners are adopting AI in the middle of a hard year, with thin staff and reasonable doubts about output quality.

What this means if you run a local business

The numbers argue for a narrow starting point. Pick one job with a clear right answer and a clear wrong one, put AI on that job, and check its work the way you would check a new hire's work during the first month. The accuracy concern that 46% of users raise is legitimate, and it shrinks when the task is bounded: a defined set of questions, approved answers, and anything outside the lines goes to a person.

Consider the phone, because the survey keeps circling back to it. Reaching customers is the top operational challenge firms report, and a missed call is the simplest version of that problem. The caller had a need and a short window. A system that answers the call, takes a message, and gathers the basic details — name, number, what the job is — then points the caller to booking, gives the owner something solid to act on in the morning. That is a modest job. It is also the kind of bounded, checkable job the data suggests AI handles best. Quotes and judgment calls stay with the owner.

This is the work we set up at Bright & Bold: AI answering for calls that come in after hours, chat agents for the questions a website gets at night, and the local search groundwork that puts a business in front of customers before they call anyone. The Fed's data cannot tell you whether any of that pays off for your shop. It does say where the payoff tends to show up first — in the places where customers were already trying to reach you. Start there, and give it a full quarter of measurement before you judge it.

About this data

Figures in this post come from the Federal Reserve Banks' 2026 Report on Employer Firms, drawn from the 2025 Small Business Credit Survey, and from a Federal Reserve Bank of New York analysis of the same survey data. The survey was fielded September 3 to November 14, 2025, and collected 6,525 responses from employer firms with 1 to 499 employees across all 50 states and the District of Columbia. The expectations analysis uses the survey's AI module, answered by 5,248 firms.

One limitation deserves a plain statement. The SBCS is a convenience sample, not a random sample. The Federal Reserve weights responses against Census benchmarks for firm age, size, industry, geography, and owner demographics, and it flags the design in its own report. Treat the figures as a carefully built snapshot of the firms that answered, not a census of every small business in the country. Adoption figures also count any use of AI by the business or its employees for work, including AI features built into software a firm already had.

Sources: Federal Reserve Banks. 2026 Report on Employer Firms: Findings from the 2025 Small Business Credit Survey. March 2026. https://doi.org/10.55350/sbcs-20260303 — AI use, depth, tasks, effects, and barriers (pp. 26–28); operational and financial challenges (p. 3). Report and findings summary: fedsmallbusiness.org/reports/survey/2026/2026-report-on-employer-firms — Aarons, Will, and Asani Sarkar. "AI Adoption and Employment Expectations: Evidence from a Survey of Small Business Owners." Liberty Street Economics, Federal Reserve Bank of New York, October 8, 2026. https://doi.org/10.59576/lse.20261008

Bright & Bold Digital Marketing, Temple City, California. Analysis of public data from the Federal Reserve Banks. The underlying report is published under a Creative Commons Attribution-NonCommercial 4.0 license; this post is our own analysis and wording.