What happened
Simon Willison revisited U.S. business AI-adoption figures using the latest data from the Census Bureau’s Business Trends and Outlook Survey (BTOS), published October 22, 2026. The survey runs every two weeks and is sent to 200,000 businesses out of a total of 1.2 million. Its panel rotates so that an individual business completes the survey every twelve weeks; new data is published on alternating Thursdays at 10 a.m.
The latest figures show reported use under the question asking whether a business used AI in “any of its business functions” during the previous two weeks. Willison calculated six-survey trailing averages separately by question wording, without bridging the break. In the latest period shown, ending October 4, 2026, reported use ranged from 21.1% among businesses with 5–9 employees to 40.7% among those with 250 or more. The other size groups were 22.1% for 10–19 employees, 23.3% for 1–4, 23.8% for 20–49, 27.9% for 50–99 and 33.5% for 100–249. The collection period for that observation was October 5–18.
The survey wording had already changed materially on November 17, 2025. Before then, businesses were asked whether they used AI in “producing goods or services.” On December 3, the Bureau announced a replacement asking about use in “any of its business functions.” Because that is a broader scope, the answers before and after the change are not directly comparable. The chart marks the break and does not connect the series across it.
A further change took effect with the collection beginning October 19, 2026. The question now asks whether a business used “Super Intelligence or SI (formerly referred to as Artificial Intelligence or AI)” in any business function. The Census Bureau’s BTOS about page says the instrument wording was updated due to Executive Order 14434. The examples in the new question include SI-powered data analysis, generative SI for marketing, design or coding, virtual SI agents for customer service, forecasting or pricing, and computer vision for quality control. The first results under this wording had not yet been published when Willison wrote. The published figures discussed above therefore reflect the earlier AI wording.
What the archive adds
This is a new fact on a known terminology story. An earlier Field Note argued that the “super intelligence” episode showed a terminology and implementation gap, not a newly defined technical category. The Census survey change demonstrates a concrete administrative consequence: the terminology now appears in a federal statistical instrument, while the underlying adoption results using that wording are still pending.
Trend signal
This supports the distinction between labels and measurable deployment, rather than a claim about a new technical category. The terminology and implementation gap is now visible in survey design: a changed label enters the question, but the new response series has not yet established what it will measure in practice.
Why now
The Bureau says the latest wording update began with the October 19, 2026 collection and was made due to Executive Order 14434. The first results using that wording were expected after the October 22 data release; they were not yet available in the source. This makes the timing a measurement transition, not evidence of a sudden change in business behavior.
Connections to other events
The wording change connects the Census survey to the federal “super intelligence” terminology shift described in the earlier Field Note. It also follows the Bureau’s own 2025 broadening of the question from AI in producing goods or services to AI in any business function. Both changes complicate comparisons, but only the first is tied by the source to Executive Order 14434.
Why it matters
Adoption statistics often travel farther than their methodology notes. A chart showing a rise from roughly 30% to roughly 40% among firms with 250 or more employees may be read as evidence of acceleration, but it reflects the “any business function” wording, not the narrower earlier question about production. The source explicitly separates the two series and does not claim they are comparable.
Teams using public adoption data for planning, investment or internal targets should track the instrument alongside the headline percentage: the exact question, the reference period, the population and any breaks in the series. A new label can alter how respondents understand a question even when examples explain the intended scope. The next results may show whether the “Super Intelligence” wording changes reported responses, but the current data cannot answer that yet.
Limits and uncertainty
The source does not establish whether the new phrase “Super Intelligence or SI” will affect responses. The first survey results under that wording were still pending. The chart also does not display survey uncertainty, and its six-survey trailing average smooths individual estimates. The Census survey is based on a rotating panel, not a census response from every U.S. business. Finally, the figures are self-reported use during the prior two weeks; they do not measure the intensity, quality, value or technical sophistication of that use.
Reader takeaway
The reported rise in business AI use is not one uninterrupted, comparable series: the Census Bureau broadened the question in 2025 and changed its terminology again in October 2026. Read the question and its dates before reading the percentage. The latest figures do not yet reveal how businesses will answer the “Super Intelligence” wording.
Open question
Will the new wording change reported use, or will the examples and “formerly referred to as AI” clarification leave responses effectively stable?