Business AI spending stalls as usage climbs, Ramp data shows

Corporate spending on artificial intelligence models has slowed down even as companies run more work through them than ever before. Figures published by corporate card and spend platform Ramp show that token volume jumped 37.1 percent in late October compared with the month before, while customer spending grew by only 8.9 percent over the same period.

The gap points to two changes across the market: falling model prices and tighter controls by business buyers. Over that four-week window, the average price per million tokens across OpenAI and Anthropic fell nearly 27 percent. OpenAI prices dropped more than 42 percent, while Anthropic prices dropped almost 16 percent.

Ramp lead economist Ara Kharazian reported that the flattening in spend followed earlier pauses through August and September, when combined growth was just 1.5 percent. Businesses are shifting workloads toward cheaper lightweight models for routine work and putting hard spending limits on staff accounts rather than letting developers call premium frontier models without oversight.

Shifting away from premium frontier models

For enterprise buyers, the need to control software invoices is reshaping daily technical decisions. According to Ramp's dataset, median businesses now spend between 12 and 15 US dollars per employee each month on AI tooling. The top 1 percent of spenders, who reached peaks close to 10,000 dollars per worker earlier in the year, trimmed their expenditure slightly.

To keep budgets steady while expanding automated tasks, companies are adopting open-weight models and prompt-routing tools. These systems inspect incoming prompts and send simple queries to small, economical models, reserving larger engines only for tasks that demand complex reasoning.

What this means for Australian businesses

Small and medium firms across Australia do not need to expand their software budgets to get more value out of AI. The market shift gives clear steps to rein in expenses while increasing automation.

• Check whether your software integrations call top-tier frontier models for basic summaries, data formatting or simple customer replies. Routing those tasks to smaller, lightweight options cuts per-token costs immediately.
• Give teams set monthly limits on paid accounts and API keys. Without usage guardrails, background scripts and unattended workflows can run up unexpected bills.
• With model providers actively cutting API rates to win market share, review your contracts and software subscriptions every quarter to make sure your provider has passed along recent price reductions.
• Focus on whether an automated workflow completes a job accurately and saves staff hours, rather than how many queries your team generates each week.