Ramp AI Index: Support Teams' Share of Corporate AI Spend Jumped From 0.8% to 13.3%
On this page 9 sections
TL;DR
Ramp's November AI Index, published Nov 11, puts US business AI spending at a new October high after growth of just 1.65% month over month. The index also broke out spend by department for the first time. Engineering teams drove 72% of classified AI spend, down from summer highs, while customer support and success went from 0.8% of classified spend in June to 13.3% in October. The average price per million tokens fell 21% in a single month. The data dropped one day after Anthropic printed IPO terms near a $1 trillion valuation, which is a bet this curve bends upward again.
What Ramp measured
Ramp issues corporate cards and pays bills for more than 85,000 US businesses, so its AI Index counts what companies pay for, not what they tell pollsters. Lead economist Ara Kharazian writes the monthly update, this one titled "New Fronts in the AI Trade."
The caveats come from Ramp itself. Token-level metrics come only from companies using its Token Spend Management product, a subsample that skews toward large AI customers, and the overall customer base skews growth-oriented and technical, so adoption is probably overstated. Free tools and anything charged to a founder's personal card are invisible, as The Decoder's writeup notes.
The department view arrived with this edition. Ramp classifies transactions into team budgets and reports shares of "classified spend," the subset it can attribute to a department. A first reading of a new dataset deserves extra skepticism, and the catches below apply.
October's growth, in context
Token spend on OpenAI grew 2.1% month over month; Anthropic grew 0.5%. OfficeChai's recap of the series shows how sharp the cooling is: OpenAI's token spend grew 53% in July, shrank 4% in August, rebounded 22% in September, then added 2.1% in October. Anthropic ran 35%, minus 13%, then 0.5%.
Prices explain part of the gap between the doom headlines and the data. The average price per million tokens fell 21% last month, and Ramp says token volume keeps growing faster than token spend. The same pattern showed up in the September index: The Decoder reported usage up about 50% since July's spending peak while dollars went sideways. Picture a gym that keeps cutting membership prices while everyone visits more often: attendance records every month, flat revenue.
One number runs against the cheap-model story. The most expensive models, Opus, Fable, Sol and Astra, took 69.5% of spend in October, up from 67.7% the month before. The flagship bill is not shrinking yet.
The bill moves past engineering
Engineering at 72% of classified spend confirms what the enterprise AI market has been so far: coding agents and the orgs that buy them. The share is down from summer highs, and the fastest-growing line is customer support and success, at 13.3% of classified spend, up from 0.8% in June. Ramp attributes the jump, with hedging in its own text, to models deployed for customer service chatbots.
Kharazian's framing: support is the second front of the AI trade, and it looks different from the first. These deployments have not matched the hypergrowth engineering posted early this year, and they tend to run on cheaper, faster models rather than flagships. For OpenAI and Anthropic, the next cohort of customers may arrive with lower margins than the coders did.
One percent writes 80% of the checks
The concentration did not budge. One percent of businesses drive 80% of revenue for OpenAI and Anthropic, and that top slice is mostly VC-backed firms in tech, finance and technical services. Outside those sectors, adoption barely spreads. The median firm spent $16.32 per employee on AI in October, up 93 cents; the top 10% spent $801, roughly flat; the top 1% spent a median of $10,200 per employee, up 9%.
The median firm's AI budget is a coffee and a pastry per person per month. The top 1% is playing a different sport, and Kharazian has flagged the risk since summer: 1% of 85,000-odd businesses is roughly 850 companies, largely correlated, whose budgets move together. The whole index rides on them.
A third vendor joins the chart
October's breakout adoption belongs to SpaceXAI, home to Grok, the Cursor coding tool and the Grokbot agent per Ramp and The Decoder: 6.3% of Ramp businesses paid it, up 2.2 percentage points, the fastest mover in the index and, per Ramp, the fastest-growing software company in its entire dataset. Anthropic adoption rose 1.1 points to 44.8% and OpenAI held flat at 40.6%. Overall, 56.8% of businesses paid for some AI tool, up 1 point, a touch faster than September's 0.66-point gain.
Kharazian expects the two-vendor price war to become a three-vendor one, which lands on per-token prices already falling 21% a month.
The catches
- The department view is one month old and covers only classified spend, the subset Ramp can attribute to a team. Treat 72% and 13.3% as a first reading, not a census.
- Token metrics come from companies using Ramp's Token Spend Management, a subsample Ramp itself calls not representative. Dollar totals cover its 85,000-plus customers, who skew young, technical and growth-oriented.
- Falling prices confound the headline. A 1.65% rise in dollars while volume keeps climbing says usage is healthier than revenue.
- One soft month is not a plateau. The index see-sawed all year, a 53% July for OpenAI made autumn look worse by comparison, and OfficeChai notes a similar flat stretch last year was followed by renewed growth into year-end.
- Ramp sees payments, not outcomes. Nothing here says whether the support-department spend earns its keep.
Why builders should care
If you build or sell AI tooling, the buyer map gained a second continent. For two years the paying customer was the engineering org; the first non-dev departments spending real money are support teams, and they buy on latency and price per ticket, not leaderboard rank. A support product priced like a flagship API loses that deal.
If you build on the models, plan for the deflation. At minus 21% a month on the blended price per million tokens, usage-based revenue has to compound hard just to stay flat, and a third well-funded vendor pressing prices down makes that harder. Growth has to come from volume, and margin from somewhere else.
And if a pitch deck near you claims enterprises are adopting AI, ask which enterprises. Adoption of 56.8% sounds broad until the same index shows 1% of companies supplying 80% of the labs' revenue. The median American business is still spending coffee money.
Key Takeaways
- Ramp's November AI Index, drawn from payments by more than 85,000 US businesses, put October AI spend at a new peak on growth of just 1.65% month over month, with OpenAI token spend up 2.1% and Anthropic up 0.5%.
- The index's first department-level dataset shows engineering at 72% of classified spend, down from summer highs, and customer support and success up from 0.8% in June to 13.3% in October, a rise Ramp links to service chatbots.
- The average price per million tokens fell 21% in a month while token volume kept growing faster than dollars, so revenue growth now understates usage.
- Concentration persists: 1% of businesses drive 80% of OpenAI and Anthropic revenue, the median firm spends $16.32 per employee per month against $10,200 for the top 1%, and adoption outside tech, finance and technical services is barely moving.
- SpaceXAI reached 6.3% of Ramp businesses, up 2.2 points, the fastest mover in the index; overall adoption reached 56.8%.
- Frontier models Opus, Fable, Sol and Astra still took 69.5% of spend, up from 67.7%, so the flagship bill is not shrinking yet.
Sources: Ramp AI Index, November 2026 (Nov 11, 2026), Ramp AI Index, The Decoder, Bastian (Nov 11, 2026), OfficeChai (Nov 12, 2026), The Decoder (Oct 2, 2026)