AI at work: adoption by the numbers

Ask what share of American businesses use artificial intelligence and the honest answer is: it depends which federal survey you ask. The Census Bureau’s Business Trends and Outlook Survey (BTOS) put firm-level adoption at 18% at the end of 2025, climbing to 19.8% nationally by early May 2026; The Atlanta Fed’s Survey of Business Uncertainty (SBU), measured the same window, put employment-weighted adoption at 78%. A worker-level tracker, the Real-Time Population Survey (RPS), landed in between at 41% of the workforce; all three are current, all three are run by credible institutions, and none of them is wrong: they are answering different questions.

Three surveys, three numbers

Why the same question gets three answers

The Federal Reserve’s own account of the gap points to three mechanics, not measurement error. First, weighting: BTOS mirrors the population of firms, while SBU is employment-weighted and skews toward larger employers; The Fed notes a core cause in its own words; quote: "the different sampling distributions combined with adoption heterogeneity across firm size classes likely drives a considerable share of the gap between the BTOS and SBU estimates". Second, question framing: BTOS originally asked about AI use in "producing goods and services," a narrower bar than the broader "any business function" wording it adopted in November 2025, a change that itself breaks strict comparability across the series. Third, who answers: SBU surveys business executives directly, and the Fed argues executives know firm-level AI usage best; its wording: "business executives that are surveyed in the SBU are in the best position to comment on a firm’s AI usage". That information-asymmetry case explains why executive-reported adoption runs higher than headline firm counts.

Size and sector decide which number applies to you

Within BTOS itself, the national 19.8% figure hides a size gradient; 37% of firms with 250 or more employees reported using AI in the period ending May 3, 2026, compared with 32% for firms with 100 to 249 employees and under 20% for firms with four or fewer employees. Two sectors ran well above the national rate as of the same date: Information at 39.7% and Finance and Insurance at 33.9%. A companion Census working paper, drawing on the same AI supplement, finds high adoption at the top: for very large firms in Information, Professional Services. And Finance, adoption reaches "50%-60% (60%-70%, employment-weighted)&quot. That is the segment where the two weighting methods converge most closely.

Where the adoption actually goes

Among firms that report using AI, the Census working paper covers the November 2025 to January 2026 supplement and finds the tool concentrated in a handful of functions. The paper names the top functions as Sales and Marketing (52%), Strategy and Business Development (45%), and IT (41%). Use is narrow even there: 57% of adopting firms use AI in three or fewer business functions. At the worker level, "in 23% (41%, employment-weighted) of firms, workers use AI in work-related tasks," mostly for writing, document analysis and information search, and 65% of those firms cap use at three or fewer tasks. On the jobs question, the same paper reports employment cuts tied to AI are rare, occurring in just 2% of firms, while 66% of adopters use AI "solely to augment tasks" rather than replace workers. Adoption is expected to keep climbing. BTOS respondents projected 22% of firms would be using AI within six months of the survey.

Methodology

Sources

U.S. Census Bureau, “Large Firms With at Least 20 Employees Biggest AI Users.” America Counts feature on BTOS data, Dec 14, 2025 to May 3, 2026. Published May 26, 2026 (census.gov/library/stories/2026/05/ai-use-businesses.html). Kathryn Bonney and co-authors (Cory Breaux, Emin Dinlersoz, Lucia Foster, John Haltiwanger, Aditya Pande). “The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks.” Census Bureau CES Working Paper CES-WP-26-25. Reference period Nov 2025 to Jan 2026, published April 2026 (census.gov/library/working-papers/2026/adrm/CES-WP-26-25.html). Jeffrey S. Allen, “Monitoring AI Adoption in the U.S. Economy,” Federal Reserve FEDS Notes, published April 3, 2026. It synthesizes BTOS, the Real-Time Population Survey and the Atlanta Fed’s Survey of Business Uncertainty, November 2025 wave (federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-s-economy-20260403.html).

Scope limits

All figures cover the United States only. No UK or EU firm-level tracker with comparable question design and cadence was located for this brief. BTOS changed its core AI question in November 2025 (from usage in “producing goods and services” to “any business function”), which breaks strict period-over-period comparability within that series. Adoption definitions are not harmonized across BTOS, RPS and SBU. The 18%/41%/78% figures answer three different questions about what counts as “using AI,” not three estimates converging on one true number.

Updates

This brief will be revised as new data is published. Corrections are handled under our corrections policy.

Updated 2026-08-03.