The 95% and the 9% Are Both Correct
Ninety-five percent of B2B marketers say their organization uses AI. Roughly 9% of US firms actually use AI in a sales or marketing function. Both figures come from 2025 and 2026 fieldwork, both are methodologically sound, and the gap between them is not a discrepancy to be resolved. It is the entire subject.
The Federal Reserve published a note in April 2026 making this point in the general case. Reviewing three high-quality surveys of US AI adoption, it found firm-level adoption at 18%, individual worker adoption at 41%, and employment-weighted firm adoption at 78%. An earlier Fed review of sixteen separate surveys found work-related AI adoption estimates ranging from about 5% to 40%. The Fed’s conclusion was not that most of these are wrong. It was that each answers a different question, and the right estimate depends on the question being asked.
Marketing statistics inherit every one of those problems and add several of their own. This article names each fracture and shows the arithmetic.
Figure 1. Seven published AI adoption rates, ordered by size.
Fracture one: firm-weighted, employment-weighted, and worker-level adoption differ by 4.3x
The Federal Reserve’s April 2026 note on monitoring AI adoption compared three surveys directly.
| Source | Date | Unit of analysis | Estimation type | Estimate |
|---|---|---|---|---|
| Census BTOS | Dec 2025 | Firms | Firm-weighted | 18% |
| Survey of Business Uncertainty | Nov 2025 | Firms | Employment-weighted | 78% |
| Real-Time Population Survey | Nov 2025 | Individuals | Share of labor force | 41% |
| Survey of Business Uncertainty (LLMs) | Nov 2025 | Firms | Employment-weighted | 54% |
Eighteen percent and 78% describe the same economy in the same month. The spread is 4.3x.
The mechanism is firm size. Roughly 95% of US firms have fewer than 50 employees, but those firms employ only 26.6% of workers. Firms with 250 or more employees are 0.9% of all firms and employ 56.2% of workers. Large firms adopt AI at much higher rates. A firm-weighted average is therefore dominated by small firms and reads low; an employment-weighted average is dominated by large firms and reads high.
Sampling compounds this. The BTOS sample is 89.6% firms with 1 to 49 employees, which closely mirrors the true firm population. The Survey of Business Uncertainty sample is 38.2% firms of that size, deliberately over-weighting large employers. Neither is wrong. They are built to answer different questions.
The Fed’s own guidance is worth stating plainly. The BTOS is the best estimate of the share of US businesses that have adopted AI. The Real-Time Population Survey is the best estimate of the share of the labor force using generative AI at work. The Survey of Business Uncertainty is a reasonable upper bound on how many workers have access to AI tools at work.
Marketing statistics almost never disclose which of these three questions they are answering.
Fracture two: only about 9% of US firms use AI in a marketing function
The most useful marketing-specific figure available comes from the Census Bureau’s second AI supplement, analysed in The Microstructure of AI Diffusion, a National Bureau of Economic Research working paper by Census Bureau economists using nationally representative data collected November 2025 to January 2026.
The findings, in sequence:
- 18% of firms used AI in at least one business function; 32% on an employment-weighted basis
- Among adopting firms, 57% use AI in three or fewer of the 15 functions tracked
- The most common function among adopters is Sales and Marketing, at 52%, followed by Strategy at 45% and IT at 41%
That produces the number almost nobody quotes:
18% × 52% = 9.4% of US firms use AI in a sales or marketing function.
Figure 2. Deriving the marketing-function adoption rate.
On an employment-weighted basis the figure is 32% × 52% = 16.6%.
Against a marketer survey reporting 95% adoption, the firm-weighted marketing figure is 10.1x smaller. Both numbers are defensible. The 95% counts marketing organisations that responded to a marketing survey. The 9.4% counts all US employer businesses, most of which have fewer than five employees and no marketing function at all.
For a B2B SaaS company deciding whether AI adoption is a differentiator or table stakes, the answer depends entirely on which population the company competes in. Among enterprise software buyers, adoption is near-universal and AI capability is table stakes. Among the long tail of small US businesses, it is neither.
Fracture three: within the marketer surveys, adoption and maturity differ by 3x
The Content Marketing Institute’s B2B Content and Marketing Trends: Insights for 2026, fielded 24 June to 14 August 2025 among 1,015 B2B marketers, is the strongest marketing-specific dataset available. Its headline is that 95% of B2B marketers say their organisations use AI-powered applications.
The same survey asked how far implementation has actually progressed.
| Implementation stage | Share |
|---|---|
| Exploratory | 20% |
| Developing | 48% |
| Established | 24% |
| Advanced | 5% |
| Leading | 3% |
Adding the top three stages gives 32% at established maturity or better. The gap between the quotable 95% and the operationally meaningful 32% is 3x, inside a single survey with a single sample.
Sixty-eight percent of respondents are exploratory or developing. CMI’s own framing is that marketers are experimenting with AI strategy while barreling ahead with AI production.
The prior year’s edition, CMI’s B2B research for 2025, fielded 25 June to 16 August 2024 among 980 B2B marketers, is not directly comparable because it asked a different question. It found 81% of B2B marketers using generative AI tools, up from 72% the previous year, with only 19% saying AI was integrated into daily processes and workflows. Fifty-four percent described their approach as ad hoc and 27% said their team did not formally use AI although individuals might.
That last category matters. A survey that counts “individual staff members may use it” as organisational adoption will produce a much higher number than one that does not. The 2026 question wording, asking whether the organisation uses AI-powered applications, sets a low bar by design.
Fracture four: rewording one question doubled the measured adoption rate
In November 2025 the Census Bureau changed the BTOS question from asking whether a business used AI “in producing goods or services” to whether it used AI “in any business function.”
The Federal Reserve Bank of Minneapolis reported the effect: the share of firms using AI in early 2026 jumped to 20%, double the roughly 10% recorded for core production use at the end of 2025.
No behaviour changed in that transition. The question changed. The measured national AI adoption rate doubled.
The Fed note quantifies the discontinuity by sector: the jump between the legacy and new series ranged from 47% for professional services, a 10.6 percentage point increase, to 159% for manufacturing, a 7.5 percentage point increase.
This is the single most important thing to know before quoting any AI adoption time series. Year-over-year growth figures that span November 2025 are measuring a definitional change, not diffusion. Any chart showing AI adoption accelerating sharply in late 2025 should be checked for this break before it is used to justify budget.
The Census Bureau flags a related caution: in the four BTOS surveys leading to year-end 2025, roughly 10% to 11% of respondents answered that they did not know whether their firm used AI.
Fracture five: any-use and daily-use are five times apart
Adoption headlines almost always report any use. Intensity is measured separately and reads very differently.
From the Real-Time Population Survey, November 2025, as reported in the Fed note:
| Frequency of work-related generative AI use | Share of workforce |
|---|---|
| Any use | 40.7% |
| At least once in the last week | 35.2% |
| Daily in the last week | 12.0% |
Daily use is 29.5% of any use. A separate mid-2025 Gallup survey put daily AI use at 8% of workers, which is 5.1x below the 40.7% any-use figure from a comparable period.
The Survey of Business Uncertainty measured duration rather than frequency and found a plurality of respondents, 35%, use AI up to one hour per week, with a further 29% using it between one and five hours.
For marketing planning this distinction is decisive. A team where every member has tried an AI tool once is not the same as a team where AI is in the daily workflow, and the published adoption statistics do not separate them unless intensity is reported explicitly.
Fracture six: adoption is measured, impact mostly is not
The CMI 2026 data shows what happens when a single survey measures both. Among B2B marketers using AI for content creation:
| Reported effect | Improved | No change | Decreased | Unsure or too soon |
|---|---|---|---|---|
| Productivity | 87% | 6% | 3% | 4% |
| Operational efficiency | 80% | 11% | 2% | 7% |
| Creative capabilities | 65% | 22% | 5% | 8% |
| Content quality | 58% | 21% | 12% | 9% |
| Content performance | 39% | 34% | 5% | 22% |
Figure 3. Reported improvement by type of effect.
The pattern is unambiguous. Effects closest to the keyboard are reported as strong improvements. Effects closest to business outcomes are not. Content performance improvement is reported by 39%, exactly 50 percentage points below the 87% productivity figure, with 34% reporting no change at all and 22% unable to say.
Twelve percent report content quality decreased.
Tool usage follows the same gradient. Eighty-nine percent use AI for content creation, but only 16% for advertising optimisation, 14% for personalisation, and 12% for predictive analytics and targeting. Adoption is concentrated in the production layer, not the decision layer.
The agentic figures show how early the category is. Twenty-eight percent of B2B marketers experiment with AI agents, rising to 43% among CMI’s self-rated pacesetters, but only 3% overall say agents are core to their strategy. Among those experimenting, 52% report improved operational efficiency, 21% better customer engagement, and 19% increased campaign performance or ROI.
The measurable pattern across every dataset reviewed is that AI adoption statistics describe input activity, not output results. Marketing teams that report AI-driven pipeline gains are reporting something the adoption surveys do not measure, which is the same reason SERPsculpt reports organic performance in pipeline and revenue terms rather than activity volume.
Fracture seven: sector variation is larger than the year-over-year trend
National averages hide a spread wide enough to make them useless for planning.
Census BTOS, as of 3 May 2026, against a national rate of 19.8%:
| Sector | Current AI use |
|---|---|
| Information | 39.7% |
| Finance and Insurance | 33.9% |
| National average | 19.8% |
| Retail Trade | ~14% |
Figure 4. AI use by sector against the national average.
The Fed note adds further BTOS figures under the new series: professional, scientific and technical services at about 33%, real estate at 24%, wholesale trade at 13%, and accommodation and food services at 8%.
At the worker level in the same period, the Real-Time Population Survey found generative AI use at 70% in Information, 63% in finance, 62% in professional services, 48% in wholesale trade, and 21% in accommodation and food services.
Two observations follow. First, the gap between Information at 39.7% and accommodation at 8% is roughly 5x, far larger than any year-over-year national movement. Second, the worker-level figure exceeds the firm-level figure in every sector, often by 20 to 30 percentage points, which is direct evidence that individual use runs ahead of formal organisational adoption. The NBER paper confirms the mechanism, finding that worker use can occur without firm adoption and firm adoption can occur without worker use.
Firm size matters as much as sector. The NBER analysis found AI use reaching 50% to 60% among very large firms in Information, professional services, and finance, rising to 60% to 70% employment-weighted. Census reports 37% for firms with 250 or more employees, 32% for firms with 100 to 249 employees, and under 20% for firms with four or fewer.
What this means for B2B SaaS marketing planning
Five questions replace the single question “how many marketers use AI.”
Which population is being counted? All US firms, firms weighted by employees, marketing organisations that answered a marketing survey, or individual workers. The answer moves the number by up to 10x.
Does the survey span November 2025? If so, part of any measured growth is a Census question rewording that doubled the national rate without any change in behaviour.
Is it any-use or intensity? Daily use runs at roughly 30% of any-use. Team capability planning should use the intensity figure.
Is it adoption or maturity? CMI’s own data separates 95% adoption from 32% established maturity. Those support very different strategic conclusions.
Is it adoption or effect? Eighty-seven percent report productivity gains and 39% report content performance gains from the same respondents. Only the second is a business result.
For a B2B SaaS company, the practical read is that AI adoption is table stakes in claim and rare in execution. Ninety-five percent will say they use it. Thirty-two percent have it established. Sixteen percent use it for advertising optimisation. Three percent run agents as core strategy. Competitive advantage sits in the gap between the first number and the last, which is a matter of operational discipline rather than tooling. That is the same reason content built to rank in Google and appear in AI answers depends on verified process rather than volume, and why retention benchmarks reveal more about a growth programme than adoption claims do.
Summary: the same question, seven answers
| Question | Answer | Source and denominator |
|---|---|---|
| Do organisations use AI? | 95% | B2B marketers surveyed, CMI 2026, n=1,015 |
| Is it established? | 32% | Same survey, established stage or better |
| Do US firms use AI? | 18% to 20% | All employer firms, firm-weighted, Census BTOS |
| Do firms use AI in marketing? | ~9.4% | 18% adopters x 52% using it in Sales and Marketing |
| Does the workforce use AI at work? | 40.7% | Individuals, Real-Time Population Survey |
| Do workers use AI daily? | 12.0% | Same survey, daily in last week |
| Do workers have access to AI at work? | 78% | Employment-weighted firms, Survey of Business Uncertainty |
Methodology and verification
Every numeric claim was verified by direct retrieval of the source document rather than an aggregator summary.
Verified by direct primary fetch
Federal Reserve FEDS Note, “Monitoring AI Adoption in the US Economy,” Jeffrey S. Allen, 3 April 2026, including Tables 1, 2 and 3 and all sector and firm-size footnotes. US Census Bureau, “Large Firms With at Least 20 Employees Biggest AI Users,” 26 May 2026. NBER Working Paper 35141, “The Microstructure of AI Diffusion,” Bonney, Breaux, Dinlersoz, Foster, Haltiwanger and Pande, April 2026, abstract retrieved in full. Content Marketing Institute B2B research for 2026 and 2025 editions, full articles retrieved including methodology sections.
Derived, with calculation shown
- Share of US firms using AI in a sales or marketing function, firm-weighted: 18% × 52% = 9.4%
- Employment-weighted equivalent: 32% × 52% = 16.6%
- Ratio of CMI adoption claim to firm-weighted marketing rate: 95 ÷ 9.4 = 10.1x
- CMI established maturity or better: 24 + 5 + 3 = 32%
- Ratio of CMI adoption to CMI maturity: 95 ÷ 32 = 3.0x
- Fed cross-survey spread: 78 ÷ 18 = 4.3x
- Daily use as share of any use: 12.0 ÷ 40.7 = 29.5%
- Gallup daily against RPS any-use: 40.7 ÷ 8 = 5.1x
- Productivity-to-performance gap: 87 − 39 = 50 percentage points
Reported secondary
The doubling of the BTOS rate following the November 2025 question rewording is reported by the Federal Reserve Bank of Minneapolis, citing Census data. Gallup’s 8% daily-use figure and the Crane, Green and Soto review of sixteen surveys are cited as reported within the Federal Reserve note rather than from the original publications.
Rejected and excluded
A widely circulated claim that 95% of B2B organisations are using or planning to use AI tools was traced to a content-farm aggregation with no retrievable underlying study and is not used, despite superficially matching the CMI figure. Claims that 88% of companies use AI in at least one function and that 78% of B2B firms apply AI across business operations appear in the same aggregation without attribution and are excluded. Vendor surveys of self-selected customer bases, including small-business AI adoption figures ranging from 8.8% to 58% depending on sponsor, were reviewed and excluded because sponsor audience composition rather than measurement drives the spread. The Census working paper CES-WP-26-25 could not be retrieved directly because the host disallows automated access; its findings are cited via the NBER version of the same research, which was retrieved in full.
Known gaps
No nationally representative survey isolates marketing-department AI adoption as a standalone population. The 9.4% figure is a derivation from two verified inputs, not a directly measured statistic, and is presented as such. CMI data is B2B marketer self-report with unavoidable self-selection: marketers who respond to content marketing surveys are more engaged with marketing practice than the average firm. The Survey of Business Uncertainty asked its AI questions once, in November 2025, with 1,032 responses, so no trend is available from that source.
Census and Federal Reserve figures are as of the collection periods stated. Content Marketing Institute figures carry fieldwork dates that precede publication by several months.
With new Ai Disclosure laws numbers are expected to shift by the end of 2026.