The U.S. Census Bureau asked roughly 200,000 businesses every two weeks, for six months, one question: are you using AI. When it published the results in May 2026, the answer split cleanly on a single number. Adoption kept climbing at every firm size above 20 employees between December and May. Below that line, it barely moved. A funded, non-technical founder running a lean team isn't failing to see the same opportunity a bigger company sees. They're on the wrong side of a headcount cutoff Census can now point to directly.
What the Business Trends and Outlook Survey actually asks
The Business Trends and Outlook Survey (BTOS) is the Census Bureau's biweekly read on U.S. employer businesses, built from a rotating panel of about 1.2 million firms split into six panels of roughly 200,000, each reporting once every 12 weeks over a year. It covers all employer businesses, single-location and multi-location, excluding farms. Starting in late 2025, the Bureau added an AI Supplement to the standard questionnaire, collected December 14, 2025 through May 3, 2026, and published the results on May 26, 2026 in a release titled "Large Firms With at Least 20 Employees Biggest AI Users," written by Census economists Adam Grundy, Cory Breaux and Dhanapati Khatiwoda. Unlike a one-off vendor poll, BTOS reports by firm size, sector, state and the largest metro areas, and the underlying tabulations are downloadable from the Bureau's BTOS data page.
Where Census draws the line
The headline finding is specific enough to quote directly: "AI use increased among firms with at least 20 employees but didn't change significantly among firms with fewer than 20 employees" over the five-month window. The size brackets Census reported by name:
| Firm size | Share currently using AI |
|---|---|
| 250 or more employees | 37% |
| 100 to 249 employees | 32% |
| 4 or fewer employees | under 20% |
Nationally, overall AI use hovered between 17% and 20% across the whole period, with 20% to 23% of businesses expecting to use it within six months. That range sounds narrow until you notice it's an average sitting on top of a bimodal reality: firms above the 20-employee line pulled the number up every reporting period, and firms below it held the average down by staying almost exactly where they started in December.
Sector doesn't rescue the smallest firms
Census also broke the same data out by industry, and the spread there is wide: the Information sector led at 39.7% current use (42% expected within six months), Finance and Insurance followed at 33.9% (39% expected), and Retail Trade sat far behind both at around 14% current use (17% expected). Worth being precise about what that comparison can and can't tell a small AI-native company: Census published sector and firm size as two separate breakdowns, not one crossed against the other, so there's no way to read off this data alone what a 12-person Information-sector firm's adoption rate looks like next to a 300-person one in the same sector. What the two breakdowns do show, read together, is that sector alone doesn't explain the gap Census flagged as tracking firm size specifically. A founder building in a high-adoption sector shouldn't assume the sector average describes their own, much smaller company.
The survey measures a function, not enthusiasm
The AI Supplement doesn't just ask yes or no. It measures use across 15 distinct business functions, including finance, human resources, customer service, marketing, information technology and research and development. That detail matters more than the headline percentages, because it reframes what the survey is actually counting. A 250-person company has someone whose job includes evaluating a new tool for one of those 15 functions and building whatever integration it needs. A 12-person company usually doesn't, not because nobody there wants AI in the product, but because there's no spare person to spend six weeks scoping a vendor and wiring it in in between the other 40 things a lean team is already doing. Census's own line, that adoption moved above 20 employees and sat still below it, reads less like a preference gap and more like a capacity one.
What this means for a funded, non-technical founder
If your team is under that line, the honest read of this data isn't that you're behind. It's that the thing standing between your company and the AI feature you already know your product needs usually isn't conviction, it's a person free to build it. That's the same gap we've written about before: growth stalling not because demand or ambition ran out, but because the systems behind the work never got the headcount to keep up. The fix isn't always the twentieth hire, which is slow, expensive, and a bet on someone you haven't met yet. A scoped build against a fixed brief closes the same gap without adding a full-time salary line, provided the brief is actually clear enough to scope, which is worth checking against a realistic MVP timeline before committing to one. The honest exclusion: if the idea is still genuinely open-ended, no defined feature, no clear success criteria, a fixed-scope partner is the wrong tool at that stage. That's a discovery conversation with your own team first, not a build.
A worked example
Picture a 14-person, seed-funded logistics marketplace connecting independent truckers to shippers. The founder has no technical background and knows customer support tickets are eating a growing share of the ops manager's week, the kind of triage an AI-assisted intake flow handles well elsewhere. Nobody on the team has evaluated a vendor for this before, and hiring an IT lead to own it would take months and a salary the seed round wasn't sized for. Under Census's finding, that's exactly the company sitting below the 20-employee line with adoption flat since December, and exactly the company for whom the constraint was never the idea. It was who was going to build it while everyone else kept the business running.
If your own company sits on the wrong side of that headcount line, the practical question this data raises isn't whether AI belongs in what you're building. It's who does the work while your team stays lean, which is the same math behind the in-house-versus-outsourced comparison we ran last year. We scope custom software builds against your actual brief, not your headcount, with transparent pricing and an itemised estimate within 48 hours of a call.
Sources
- U.S. Census Bureau: Large Firms With at Least 20 Employees Biggest AI Users (published May 26, 2026, by Adam Grundy, Cory Breaux and Dhanapati Khatiwoda; accessed September 25, 2026)
- U.S. Census Bureau: Business Trends and Outlook Survey data (methodology and downloadable tabulations by firm size, sector, state and metro area; accessed September 25, 2026)
- U.S. Census Bureau: Business Trends and Outlook Survey (BTOS) program page (panel structure and sample coverage; accessed September 25, 2026)