37% of Large Firms Use AI. Under 20% of Four-Person Firms Do — and That Number Stopped Moving.

The Census Bureau has been asking American businesses every two weeks whether they used AI. Between December and May the answer went up for firms with at least 20 employees and did not move for firms with fewer. The adoption gap is not closing. It is widening, and the reason has almost nothing to do with the software.

The number is a size story, not a technology story

Every founder has now been told that AI changes everything. The Census Bureau has been quietly checking whether it has. Since 2023 the Business Trends and Outlook Survey has asked a nationally representative sample of American firms, every two weeks, whether they used artificial intelligence. It is the closest thing there is to a live read on adoption, and it is not a vendor survey — nobody in the chain is selling AI.

The most recent six months ran from 14 December 2025 to 3 May 2026. Nationally, use hovered between 17% and 20%, finishing at 19.8%. That is the number that gets quoted. It is also the least useful number in the release, because it averages together a 250-person company and a two-person company as if the question meant the same thing to both.

Broken out by size, the picture separates. Thirty-seven percent of firms with at least 250 employees reported using AI. Thirty-two percent of firms with 100 to 249 employees did. For firms with four or fewer employees, the Bureau does not publish a rate at all — it publishes a ceiling: less than 20%.

The gap did not hold steady. It opened.

A static gap between large and small firms would be unremarkable and, frankly, survivable. Big companies get to things first; small ones catch up when the tooling gets cheap and obvious. That is the ordinary shape of technology diffusion and it has happened with accounting software, with e-commerce, with payments.

That is not what the Census data show. Over the six months to May, AI use increased among firms with at least 20 employees. Among firms with fewer than 20 employees it did not change significantly. One end of the distribution moved and the other end sat still. Whatever was happening in the market during those six months — and a great deal was — it reached firms above roughly twenty people and stopped there.

Six months in which AI was the only thing anyone talked about produced no measurable change in whether the smallest firms were using it.

Sector explains less than you would expect

The obvious objection is that this is really an industry story wearing a size costume — that small firms are concentrated in trades and retail where AI has less to offer, and large firms cluster in information and finance where it has more. There is something to that. Information sat at 39.7% and Finance and Insurance at 33.9%, both well above the 19.8% national rate. Retail Trade ran around 14%, below it.

But sector does not do the work people assume. Retail at roughly 14% is five points below the national average; firms with 250+ employees are seventeen points above it. Size moves the number further than industry does. And the sector rates barely shifted over the six months either — the Bureau found no statistically significant movement in these sectors’ current or expected use. The thing that moved was the large end of the size distribution.

What the smallest firms are actually short of

Here is the part the adoption statistics cannot show you, so we will say it plainly. The barrier at the four-person firm is not licensing cost. The tools that matter are twenty or thirty dollars a month and several of the useful ones are free. Price is not what is holding the line at under 20%.

What a four-person firm does not have is the hour. Adopting AI into a business function — which is what the survey asks about, not whether someone drafted an email with it — means somebody has to choose the function, test the tool against real work, decide it is good enough, rewrite the process around it, and teach the other three people. In a 250-person firm that is somebody’s job for a fortnight. In a four-person firm it is the owner’s evening, competing against invoicing, a customer who has gone quiet, and payroll on Friday.

That is why the gap widened rather than closed. Adoption is not gated on the technology getting cheaper or better, both of which happened during these six months. It is gated on somebody having uninterrupted attention to spend, and that is the one input the smallest firms have least of. The founders who did move on AI this year almost all did it the same way: not alone, and not by reading more about it.

One more thing worth reading carefully. Expected use ran 20% to 23% against current use of 17% to 20% — a gap of roughly three points, all year. Businesses are not predicting a wave. They are predicting a slightly larger version of now. When the people running the firms are asked directly, the honest forecast is incremental, and it has been incremental for six months straight.

Published by The Lonely Entrepreneur — the community and coaching platform for entrepreneurs who are building alone. lonelyentrepreneur.com

This article is for educational purposes and is not a substitute for professional financial or legal advice.