Four separate federal and bank-affiliated reports released this spring describe the same American economy, and they describe it the same way: a two-track A.I. rollout in which firms with 250 or more employees are pulling decisively ahead of everyone else.

The Census Bureau’s Business Trends and Outlook Survey, drawing on data collected between December 14, 2025 and May 3, 2026, found 37 percent of firms with at least 250 employees reported using A.I., and 32 percent of firms with 100 to 249 employees. Among firms with four or fewer employees, the figure sits below 20 percent. Overall adoption across the economy has hovered between 17 and 20 percent. Use, the Census Bureau noted in its report, “increased among firms with at least 20 employees but didn’t change significantly among firms with fewer than 20 employees” over the six-month window.

Weight the same data by employment rather than firm count and the headline number jumps to 32 percent for the November-to-January reference period, according to Census working paper CES-WP-26-25. In Information, Professional Services, and Finance, very large firms cluster between 60 and 70 percent adoption. The depth of use, though, is shallow: “57 percent of users integrate AI in three or fewer business functions,” the working paper’s authors write, with sales and marketing dominating.

The Federal Reserve’s April 3, 2026 FEDS Note arrives at a similar place from a different angle, pegging year-end 2025 adoption at roughly 18 percent and observing that “the first four surveys of the new BTOS series show a stronger association between adoption and firm size.” Whether A.I. eventually becomes “an equalizer of sorts,” the Fed’s staff economists concede, remains an open question.

It isn’t acting like one yet. The JPMorganChase Institute, working from transaction data rather than self-reports, finds that “Employer firms consistently outpace nonemployers regardless of revenue,” and that the gulf “widens beginning in 2023, as larger firms experience a more rapid acceleration in AI uptake.” The Institute attributes the lag to capital constraints, limited technical expertise, and integration costs. The Federal Reserve Bank of Minneapolis, citing a 30 percent versus 17 percent split between larger and smaller firms, calls the divide “one of the clearest structural findings of the current data cycle.”

A cohort of vendors has noticed. Salesforce’s small-business tier, HubSpot’s Breeze, and LemonLime, a model-agnostic deployment layer aimed squarely at small and mid-size firms, are all pitching the underserved end of the curve. Whether any of them moves the federal numbers is the question the next BTOS round will begin to answer.

For now, the data describe something closer to the cable-build-out of the late 1990s than to a general-purpose technology diffusing evenly: the firms with the balance sheets to wire themselves up are wiring themselves up first.

Sources