AI Adoption Is Accelerating — But the Real Story Is the Industry Divide
More businesses are paying for AI services. The macroeconomic question is whether expanding access translates into measurable productivity growth — and why industries are progressing at very different speeds.
In September 2026, the Ramp AI Index shows that 55.9% of businesses in its sample were paying for AI products or services. The aggregate conceals major structural differences.
1. A broad acceleration, but not a uniform one
Information and finance were early leaders, but the spread of paid AI products since 2023 has extended into manufacturing, retail, healthcare, construction and hospitality. The paths differ significantly: information businesses are near 80%, while accommodation and food services remain around one-third.
2. The September 2026 industry divide
The latest readings range from 79.9% in information to 33.0% in accommodation and food, a gap approaching 47 percentage points. Finance and insurance stands at 73.7%; manufacturing at 60.9%, ahead of the 55.9% sample-wide reading.
3. Why manufacturing matters
Manufacturing's position is notable because AI can affect physical-capital utilization as well as office workflows. Potential applications include predictive maintenance, production scheduling, quality control and inventory coordination. The investment implications depend on whether these applications deliver measurable efficiency rather than merely adding software expenditure.
4. Adoption is not the same as productivity
Paying for an AI product is an indicator of commercial adoption, not its intensity, effectiveness or economic return. Meaningful changes in labor productivity, total factor productivity and unit costs depend on organizational redesign, data integration, complementary capital spending and workforce adaptation. Benefits may emerge with substantial lags.
5. The macroeconomic consequences
Uneven diffusion can widen cost and competitive differences between sectors. In information-intensive activities, AI may compress the time required for knowledge work. In manufacturing, it could eventually increase utilization and reduce downtime. In physical and hospitality services, automated solutions face different constraints. None of those potential channels can be identified causally from adoption rates alone.
Data and methodological note
The Ramp AI Index tracks observed AI spending among Ramp customers. This is not a representative census of all US firms, and the index does not directly measure usage intensity or realized productivity. The supplied chart provides September 2026 data; the first figure reproduces its trend visually using approximate intermediate points, while the second preserves the supplied industry percentages.
Source: Ramp AI Index via Federal Reserve Bank of St. Louis (FRED). Ramp AI series on FRED.