Ramp AI Index Finds US Firms Using More AI but Spending Less on It
Ramp's latest AI Index shows corporate AI adoption rising even as per-company spending falls, a sign that inference costs are dropping faster than usage grows.
US businesses are adopting artificial intelligence at a faster clip than a year ago, yet the money they spend on it is shrinking, according to the latest Ramp AI Index compiled by Ramp economist Ara Kharazian. The finding cuts against the assumption that wider AI use automatically means bigger corporate bills, and it suggests the economics of the technology are shifting under companies' feet.
The index tracks spending across Ramp's customer base, which gives it a real-time read on how mid-market and enterprise finance teams are allocating budget rather than relying on survey sentiment. Kharazian's takeaway is that usage and expenditure are moving in opposite directions: more employees and departments are touching AI tools, but the average dollar outlay per firm is declining. That combination points to cheaper model calls, more aggressive vendor discounting, or both.
Falling per-unit costs are the most plausible driver. Model providers have spent two years slashing prices for equivalent capability, and open-weight alternatives have given buyers leverage they did not have when the first wave of enterprise contracts was signed. A company that paid premium rates for a fixed volume of tokens in 2023 can now get comparable output for a fraction of the price, so it can expand usage without expanding budget.
The shift matters for the companies selling AI infrastructure. If corporate wallets are not growing in step with adoption, revenue growth has to come from volume, new use cases, or moving up the stack into higher-margin applications and services. Vendors that priced on the assumption of ever-rising seat counts may find finance departments pushing back harder at renewal time, especially as procurement teams get better at benchmarking what they pay per task.
There is also a strategic angle for the buyers. Cheaper AI makes it easier to justify embedding the technology in routine workflows, from document review to customer support triage, because the payback period shortens. Ramp's data implies the constraint on enterprise AI is no longer licensing cost but organizational capacity: figuring out where the tools actually help and getting employees to use them consistently.
The open question is whether the spending decline is a temporary dip caused by price competition or a durable pattern in which AI becomes a cheap commodity input rather than a premium line item. Ramp's next few index updates will show whether adoption keeps climbing while budgets stay flat, which would be good news for users and a harder story for vendors banking on expanding contracts.