Top AI spenders cut per-employee costs by nearly 10 percent in August

What happened

New data from Ramp's AI Index shows that the top 1% of US companies by AI spending cut their per-employee AI costs by close to 10% in August, even as their overall use of AI tools continued to grow. The decline tracks a sharp fall in the price of AI model tokens, which Ramp's figures put at 41% since March.

The index, compiled from Ramp's corporate card and expense data, suggests that heavy AI adopters are becoming more efficient buyers of AI capability rather than simply spending more. As foundation model providers compete on price and efficiency, the cost of running AI workloads is falling even for the businesses using them most intensively.

Why it matters

This is a signal about the economics underpinning enterprise AI adoption, not just a pricing footnote. As token costs fall, the unit economics of deploying AI across customer service, operations and back-office functions improve, which should widen the business case for scaling AI beyond pilot projects. Organisations that have been cautious about the cost of large-scale AI deployment may find the calculus shifting faster than expected.

For leaders planning digital transformation and AI investment, the trend suggests that the constraint on scaling AI is increasingly less about raw compute cost and more about integration, governance and how well AI is embedded into workflows and experience design. Falling per-unit costs make it easier to justify broader rollouts, but they also raise the bar: cheaper AI removes cost as an excuse for slow or shallow adoption.

By the numbers

• Nearly 10% — the drop in per-employee AI spend among the top 1% of US firms by AI spending, recorded in August, according to Ramp's AI Index.
• 41% — the fall in AI token prices since March, per the same data.
• Top 1% — the segment of US companies tracked, defined by the highest overall AI spend.

The Renascence take

Falling AI costs are usually framed as a straightforward efficiency win, but the more interesting question is what heavy adopters do with the savings.

Cheaper tokens don't automatically translate into better experiences — they just lower the barrier to trying more things. The organisations that benefit most won't be the ones that pocket the savings, but the ones that reinvest the freed-up budget into rethinking how AI actually touches customers and employees: fewer isolated pilots, more AI embedded directly into service journeys where it changes waiting times, resolution rates or the quality of a decision. The real risk with falling costs is complacency — treating cheaper AI as a reason to do more of the same, rather than a reason to redesign the experience itself.