Ramp says AI token spend grew just 1.5% across August and September while usage kept climbing. Route cheap work to cheap models

Businesses are using more AI than ever, but paying about the same for it. The best spenders got there on purpose.

What happened

TechCrunch reported on new Ramp data showing that token spend grew only 1.5% across August and September, compared with 42% month over month growth in July, even as token volumes rose. October picked up again, with token spend up 8.9% from September 23 to October 23. In its October AI Index, Ramp said the top 1% of firms spent $9,568 per employee per month in September, down 3.1% from a July high of $9,873. Per TechCrunch, median firms spend $12 to $15 per employee per month and the top 10% spend $600 to $1,000.

Ramp's lead economist Ara Kharazian points to price cuts from OpenAI and Anthropic, plus companies getting thriftier. He told TechCrunch that advanced users set defaults that keep expensive models off long running tasks, put budgets on employees, and push work from standard models to cheaper ones through routers. "For a lot of AI use cases, you just need a simple Yes/No," he said. Ramp also wrote that the blended price per million tokens is falling.

My take

Ignore the investor angle. The useful part is what the thriftiest companies do, because it is exactly how client automations should be built.

Most steps in a business workflow are small decisions: is this lead qualified, which pipeline does this ticket go to, is this invoice a duplicate. Those do not need a frontier model. A cheap model, or a fixed list of allowed answers, handles them for a fraction of the cost.

How I would apply it:

1. Tag every AI step by job: classify, extract, write or reason.
2. Send classify and extract steps to the cheapest model that passes your test set.
3. Keep frontier models for the few steps that need real judgment.
4. Set a monthly budget per workflow and alert before it is hit.
5. Re-check model choice every quarter, because prices keep moving.

At the median $12 to $15 per employee, waste is small. At scale, routing is the difference between a workflow that pays for itself and one that gets cut.