Business Trends and Outlook Survey
US Census Bureau firm-weighted AI adoption survey; produced 18% adoption rate for late 2025 — the lowest of three conflicting federal measures.
Last refreshed: 16 April 2026 · Appears in 1 active topic
Why does the government's own firm survey show only 18% AI adoption when workers report much higher?
Timeline for Business Trends and Outlook Survey
Fed admits: no single AI adoption figure
AI: Jobs, Power & Money- Why does the US government show 18% AI adoption when other surveys show 78%?
- The 18% figure comes from BTOS, which measures firms (firm adopts AI: yes/no). The 78% figure comes from SBU, which weights by employment — so a large AI-using company counts for many workers. The same economy, three incompatible answers.Source: Federal Reserve Board (FEDS Notes)
- What is the Business Trends and Outlook Survey?
- BTOS is a US Census Bureau and BLS rapid-response survey measuring firm-level economic conditions. Its late-2025 AI adoption finding of 18% was the lowest in a Federal Reserve reconciliation comparing three federal instruments.Source: Federal Reserve Board
Background
The Business Trends and Outlook Survey (BTOS) is a rapid-response survey run jointly by the US Census Bureau and the Bureau of Labor Statistics, measuring economic conditions at the firm level. In its late-2025 wave, BTOS recorded 18% AI adoption among US businesses — the lowest figure of the three federal instruments compared in the Federal Reserve Board's reconciliation paper published 3 April 2026. The 18% figure is firm-weighted: it measures what share of firms have adopted AI, without adjusting for how many workers are employed at those firms.
The Fed reconciliation paper showed the same late-2025 economy could be described as 18% (BTOS), 41% (RPS, individual self-report) or 78% (SBU, employment-weighted) AI-adopted, a 4.3x range from the same statistical quarter. Daily AI use from the same sources sat at 12% of the workforce; weekly use at 35.2%. The 18% BTOS figure is the most conservative because it counts a firm as either AI-adopting or not, regardless of scale. A single-employee consultancy and a 10,000-person manufacturer count equally.
The policy consequence of the 4.3x spread is that Congress cannot legislate against a single baseline; the Hawley-Warner Coalition's request for one number has received three. BTOS's 18% represents the floor of the federal measurement range, and because it undercounts large employers' workforce exposure, it is likely the most misleading of the three for labour-market policymaking.