Customer service teams went from under 1% to 13% of tracked company AI spending in four months, Ramp data shows

Customer service teams are quickly becoming a major buyer of AI inside American companies. In June, support teams accounted for 0.8% of the AI spending that Ramp, a corporate card and expense company, could assign to a department. By October, their share was 13.3%.

That shift stands out because most of Ramp's latest data, published November 11, points to slowing growth. Total business AI spending reached a record in October but rose only 1.65% from September. For AI vendors heading toward public markets, the question is shifting from whether companies buy AI to which teams will keep the bills growing.

Where the new money is going

Ramp's AI Index tracks card and bill payments at more than 85,000 US businesses. In its latest report it began sorting AI spending by the department that pays for it. Engineering still dominates, at 72% of classified spending, though engineering spending has fallen from its summer highs.

Support is the clear mover. Ara Kharazian, Ramp's lead economist, says support teams are driving a modest pickup in AI spending. He suggests customer service chatbots as a possible cause. Ramp's data records payments, though, not what companies built with them, so the chatbot explanation remains his reading.

For executives deciding where AI pays off, the support number is worth watching. Support is one of the few corporate functions measured in tickets, response times and headcount. That makes savings easier to count than in engineering, where faster coding does not always show up on the income statement.

Adoption is still widening, slowly. In October, 56.8% of businesses on Ramp paid for an AI product, up one point in a month and 13.3 points in a year. Anthropic's paid adoption rate rose 1.1 points to 44.8%. OpenAI's held at about 40.6%.

A third vendor is gaining. SpaceXAI (listed as xAI on Ramp's index page) reached 6.3% of Ramp businesses, up 2.2 points. Kharazian expects the two-way race to become a three-way one, which would add price pressure on the incumbents.

A few big spenders still carry the market

Spending remains extremely lopsided. The median firm that pays for AI spends $16.32 per employee per month, up 93 cents. Firms in the top 10% spend about $801, roughly flat. The top 1% spend a median of about $10,200 per employee, up 9%.

Kharazian estimates that this top 1% supplies roughly 80% of the business revenue OpenAI and Anthropic earn from Ramp's customers. Most are venture-backed firms in tech, finance and technical services, and their spending tends to move together. He told TechCrunch last week that almost all AI spending comes from "venture-backed, high-growth tech companies."

Vendor growth was slow. Business spending on OpenAI's models rose 2.1% in October, according to Ramp. Spending on Anthropic's rose just 0.5%. For comparison, token spending jumped 42% in July alone, then grew only 1.5% across August and September combined, according to Ramp figures reported by TechCrunch.

Kharazian also argues that most businesses may never need to spend like the top tier. A doctor's office or a car dealership, he told TechCrunch, does not necessarily need that much AI. If he is right, broader adoption will not close the gap between the median firm and the top 1%.

Cheaper AI, smaller bills

Falling prices are part of the slowdown. The average price per million tokens, the units of text AI models process and charge by, fell 21% in a month. Among firms using Ramp's token-tracking service, a smaller group than the main index, usage rose 29% in the four weeks to November 1 while spending rose only 10.5%. "You can use more AI, but use it for less," Kharazian told TechCrunch.

The rise of support could deepen that pattern. Kharazian argues that customer service work is a good fit for faster, cheaper models. For now that is a forecast, not a finding. In October, the most expensive frontier models actually took a larger share of spending, 69.5%, up from 67.7%.

The data has limits. The figures describe businesses that use Ramp, not the whole economy. Ramp itself says it likely undercounts AI use, because free tools leave no payment record. And the department breakdown is new, covering only the spending Ramp could match to a team.

Even so, the direction is useful for buyers and investors. The next leg of business AI growth may come from cost-cutting functions like support, not from engineers using ever more AI. It may also arrive at lower prices per unit. That helps corporate budgets, and makes vendors' revenue forecasts harder to hit.