Our homepage cites four numbers about AI and small and medium businesses. This article is where those numbers live: each one with its source, its reference period, its population — and, just as honestly, its caveats. Figures are attributed throughout; the interpretation is ours and is labelled as such.

~400 million SMEs worldwide

Source: World Economic Forum SME Resource Hub (global estimate). WEF SME Resource Hub ↗

This is an attributed estimate of the global SME population, not a census count. It is useful for one thing only: the scale of the audience. Whatever percentage of SMEs adopt anything, the absolute numbers are enormous — which is why a one-point shift in adoption represents millions of businesses. Treat any finer claim built on this figure with suspicion; treat the scale itself as solid.

≈20% of EU enterprises use AI

Source: Eurostat news release, December 2025. Enterprises with 10+ employed persons, EU, 2025. Eurostat release ↗

The headline average hides the story. Adoption stratifies sharply by size:

  • Small enterprises (10–49 staff): 17% — ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇
  • Medium-sized (50–249): 30.4% — ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇
  • Large (250+): 55% — ▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇

Reading (ours, not Eurostat's): larger organisations adopt first because integration capacity — data infrastructure, people who own the change — comes before tooling. That is precisely the capacity a FORT-style engagement is designed to lend.

31% of SMEs use generative AI — but mostly at the edges

Source: OECD, Generative AI and the SME Workforce (2024 publication). Representative survey, 5,000+ SMEs, seven countries. OECD publication ↗

Two numbers travel together here. Thirty-one percent of surveyed SMEs used generative AI — and among those users, only 29% reported using it in core business activities. The second figure is the one worth remembering: even where generative AI has arrived, it has mostly arrived at the edges — drafting, ideation, peripheral tasks — rather than inside the workflows the business is paid for.

Rising, uneven: adoption versus integration

Source: OECD, Empowering SMEs in the Age of AI, April 2026. Sample of 2,000+ SMEs across 12 countries — non-representative, so read it as direction of travel, not prevalence. OECD publication ↗

The consistent finding across studies through 2026: adoption keeps rising while strategic and secure integration lags behind. Businesses are using AI somewhere; far fewer have integrated it into how the business runs. We wrote the longer version of this argument as the closing essay of our AI retrospective — The AI implementation gap.

What we take from it (our reading)

  • Adoption is not the same as capability. The gap between buying tools and operating with AI is where most programmes stall — and it is an engineering problem, not a procurement one.
  • Size predicts adoption because size proxies integration capacity. Close the capacity gap and the adoption gap follows; that is the fractional-model argument in one line.
  • Caveats are part of the number. An estimate is an estimate; a non-representative sample shows direction, not prevalence. Anyone quoting these figures without the caveats is marketing, not measuring.

If you want this discipline applied to your own numbers — baselines, KPIs that fire, monitoring mechanisms that trigger action — that is our reporting, analytics and monitoring service. And if you want to know where AI can safely meet your actual operations, start with a technical review.