The owner of a fifteen-person accounting firm has been reading the same headline number across half a dozen articles in the past month. Eight percent of US small businesses have AI in operations. She has been quietly concluding that her firm is not behind, because she is roughly at that level herself. She uses an AI writing assistant for some client emails. The team is starting to use a smart-search tool over their document archive. By the all-business benchmark, she is at the median.
The benchmark she is using is the wrong one. Her sector’s actual operational AI rate, according to Federal Reserve data, is closer to thirty percent. By the accounting profession’s standard, her firm sits well below the median, even though it sits at the all-business median. She has been comparing herself to the entire small-business universe when the relevant cohort is professional services. The conclusion she has been reaching is wrong, and the urgency calculation that follows from it is wrong.
This is the most common reference-point mistake in AI for owner-managed firms, and it cuts both ways. Some founders feel behind when they are at sector baseline. Others feel comfortable when they are well below it. The fix is the same. Stop using the all-business average. Look up your sector. The difference, in most cases, is large enough to change the conversation.
Why is the 8.8 percent figure the wrong reference point?
The Census Bureau’s BTOS data on operational AI usage produces an all-business average of 8.8 percent. The figure is widely cited because it’s rigorously measured (operational use of AI in production of goods or services, not self-reported general usage) and because it gives a clean single number for headline writers. It is, for the same reasons, the least useful benchmark for any specific firm.
Operational AI adoption varies enormously by sector. Industries with knowledge-intensive work and strong customer-data infrastructure have integrated AI faster than industries with physical product workflows or deeply analogue operations. Averaging across all of them tells you about the macro picture. It tells you very little about any individual firm’s competitive position.
The sector-level numbers are also more reliable than they look. The Federal Reserve has been tracking AI adoption quarterly through BTOS since 2023, with sector cuts that hold up across multiple measurement methodologies. The professional services and financial services figures sit consistently in the 28 to 33 percent range. The education services figure sits consistently around 9 percent. The variance across sectors is genuine, with the gap between sectors larger than the gap between any sector and the headline average.
What does the sector breakdown actually show?
Federal Reserve BTOS data on operational AI usage shows the highest adoption sectors clustering around 30 percent, well above the all-business baseline. Information Technology sits at 18.1 percent. Professional, scientific, and technical services run as high as 33 percent on some measurements. Financial services sit around 30 percent. Educational services come in at 9.1 percent. The spread across sectors is much larger than the gap between any sector and the headline average.
Function-level data sharpens the picture further. Across all businesses using AI, Customer Service runs at 56 percent, Cybersecurity and Fraud at 51 percent, CRM at 46 percent, Content Creation at 35 percent, Accounting at 30 percent, and Recruiting at 26 percent. The work the AI is doing matters for the right benchmark. An accounting firm doing nothing with AI is benchmarking against a sector where AI in accounting work is at thirty percent. A recruiting firm at the same activity level is benchmarking against twenty-six percent. The baseline is a function-and-sector matrix, not a single number.
The UK numbers from the British Chambers of Commerce 2025 research show the same pattern at smaller absolute scale. B2B services (finance, law, marketing) sit at 46 percent, while B2C and manufacturing sit at 26 percent. IT and telecoms lead at 56 percent. Marketing at 53 percent. The sectoral split is consistent across the Atlantic.
What changes when you swap reference points?
Two things, mostly. The first is the urgency calculation. A founder at zero percent operational AI in a sector at thirty percent is materially behind. A founder at zero percent in a sector at nine percent has time, and may be making the right call by waiting for second-mover advantage. The action implied by “behind” depends on which sector you are in.
The second is the competitive scope. Knowing your sector’s baseline tells you what your peer firms are likely already doing. Sectors at thirty percent operational adoption are sectors where AI is in customer service, CRM, content production, and document workflows, because that is where the function-level data points. The firm thinking about whether to act is competing with peer firms who are already, in many cases, using AI to compress those functions. The right benchmark turns “should we be doing AI” into “where are we relative to where our peers are already”.
The third effect, smaller but useful, is the conversation gets specific. “Are we behind on AI” is not an answerable question. “Are we behind on AI given that financial services operational AI adoption is around thirty percent and we are at zero” is much sharper. The number is the prompt for the next question, which is which functions to act on first.
What does the trajectory tell you?
The most striking finding from the US Chamber of Commerce 2025 research is that the gap between large and small firms on AI adoption shrank from 1.8x to 1.2x in a single year. Small business adoption is accelerating faster than large enterprise adoption, and the sector averages are moving with it. UK SME usage rose from 25 percent in 2024 to 54 percent by early 2026. Today’s “behind” will look further behind by year-end.
This matters for the decision the reader is sitting with. The cost of waiting another year is not the same as the cost of waiting last year. The peer benchmarks are rising. A firm that was at sector median twelve months ago, and has done nothing since, has slipped relative to peers without doing anything wrong. The default move is no longer the safe move.
The trajectory does not justify panic. It justifies a serious look at your own sector’s number, your function-level exposure, and a realistic plan for where to act first. Acting against the right benchmark produces forward motion. Acting against the wrong one produces failed engagements (a topic covered in a separate post on this site).
What is the practical move?
Three steps, in this order. First, look up your sector’s actual operational AI adoption rate. Federal Reserve BTOS data, ad-hoc industry surveys, and trade body publications all carry the figures. Get the number that applies to your specific industry rather than the cross-industry average.
Second, look up the function-level numbers for the work your firm actually does. Customer service. CRM. Content production. Document workflows. Accounting. Recruiting. The function rate is a better predictor of where you can move quickly than the sector rate, because it tells you what specific work AI is already doing in firms like yours.
Third, place your firm honestly on both axes. Sector baseline. Function-level baseline. Where are you above, level, or below? In most owner-managed services firms, the answer is “below sector average on most functions, well below on a few, at par on one or two”. That mapping is the start of an actual plan, replacing the all-business-average sense of vague unease with something specific.
If you’d like to walk through that mapping for your firm specifically, book a conversation.
