A survey finds that while 50% of businesses are using AI in 2022, 2.5x higher than in 2017, AI adoption has plateaued between 50% and 60% for the past few years
Context & Ripple Effects
McKinsey's 2022 finding lands mid-arc in a decade-long adoption story that earlier coverage measured very differently: US Census Bureau data on 583,000 firms showed just 8.9% of businesses used any form of AI in 2018, so reaching 50% by 2022 was a genuine acceleration — 2.5x the 2017 level. The new wrinkle is the plateau: adoption has been stuck between 50% and 60% for years.
That stall sets up the tension the rest of the corpus explores. The same period saw surveys find only 35% of execs saying their org pursues transparent, accountable AI use, suggesting many adopters were checking a box rather than building capability — which is exactly where the later ROI and usage numbers pick up.
First-order effects
- For the roughly half of businesses still outside the adopting cohort, the easy wins are gone: McKinsey's plateau means the marginal firm now faces the harder integration cases, not the low-hanging pilots that drove the 2017-to-2022 doubling.
- Consultancies like McKinsey lose their clearest sales narrative — 'adoption is accelerating' no longer holds, so advisory work shifts from convincing firms to start toward fixing stalled programs.
Second-order effects
- Vendors and consultancies are pushed from breadth metrics to depth metrics: with firm-level adoption flat, the battleground becomes whether initiatives deliver returns — an IBM survey of 2,000 CEOs found only 25% of AI initiatives delivered expected ROI, giving skeptics inside stalled adopters hard ammunition.
- Employee-level usage diverges from firm-level adoption: Gallup tracked workplace AI use rising from 21% of US employees in 2023 to 40% in 2025, meaning individual workers pull AI in through tools regardless of whether their employer counts as an 'adopter'.
Third-order effects
- If the pattern holds, the industry's scoreboard changes structurally: headline adoption percentages stop being the KPI, replaced by measures of scaled deployment and realized value — the gap between 50-60% adoption and single-digit enterprise-wide scaling becomes the defining metric of enterprise AI maturity.
- Adoption may resume not through boardroom decisions but bottom-up, as employee-driven usage normalizes AI the way earlier workplace software spread — flipping the causal arrow from 'firms adopt, employees follow' to 'employees adopt, firms formalize'.
The trend: Enterprise AI is moving from a race to adopt — which plateaued around half of businesses — to a race to extract measurable value from adoption already banked.