Companies are spending more on AI than they can easily track. What started as a handful of API calls and token-based charges has turned into a fast-moving cost category that finance teams are struggling to keep on top of.
This is a problem that Ramp, the $32 billion fintech giant, is targeting with a new product focused squarely on AI spend visibility. The New York-based company, best known for corporate cards and expense management, says that it’s pulling token-level usage data from AI providers directly into its platform to give finance teams a clearer picture of where their money is going.
A growing cost with limited visibility
AI is no longer a small, experimental budget line. It’s emerging as one of the fastest-growing areas of corporate spend, one without the same level of oversight as other categories.
Ramp, for its part, has also been leaning into AI internally. The company has built systems that can generate and test code as part of its own development process, giving it a front-row view of how quickly usage, and costs, can spiral. That became clear when its own finance team tried to quantify what was being spent.
Karim Atiyeh, Ramp’s co-founder and CTO, says in a blog post published Thursday that the issue emerged when his team attempted to answer a basic question about internal AI costs.
“Last quarter, I asked our Finance team a simple question: how much are we spending on AI and where is it going? The analysis took days and still couldn’t provide the level of detail I wanted.”
Part of the problem is how AI usage is billed. Unlike traditional software contracts with fixed pricing, AI costs are tied to usage – often measured in tokens – and can shift quickly depending on how models are used.
“AI inference behaves nothing like the spend categories finance teams are used to managing,” Atiyeh explained. “It’s usage-based and volatile.”
Ramp says its internal data shows that average monthly token spend has increased 13-fold since January 2025. At the same time, costs among heavy users can jump by 50% or more within a single quarter.
Businesses’ API costs are surging (Credit: Ramp)
That volatility makes it difficult for finance teams to categorize spending or forecast future costs.
Moreover, it creates a disconnect between engineering teams, who are often closest to usage, and finance teams responsible for budgeting and reporting. Atiyeh said existing tools surface usage data, but often fail to translate it into something finance teams can act on.
“It doesn’t tell the controller whether a given API key’s spend belongs in COGS [cost of goods sold] or OpEx [operating expenses],” he says.
Bringing financial context to AI usage
Ramp’s approach is to combine billing data with usage data in a single system. By integrating directly with providers such as OpenAI and Anthropic, as well as model gateways like OpenRouter, the platform can track how tokens are being consumed across teams and projects.
The idea is to move beyond raw usage metrics and attach meaningful financial context to them.
That means tracking total spend alongside token usage, average cost per day, and cost per request, while also breaking down spending by provider, model, team, and even individual users. Finance teams can see which products or features are driving costs, and classify that spend into standard accounting categories such as cost of goods sold or operating expenses.
Ramp’s AI spend dashboard (Credit: Ramp)
Atiyeh says the issue for most companies so far hasn’t been a lack of data, but the absence of a clear way to interpret it.
“The missing layer is financial context. AI is on track to become one of the largest cost centers in business. For some companies, it’s already exceeding what they spend on payroll.”
Ramp’s product is built around a set of controls aimed at finance teams rather than developers. Beyond visibility, it allows teams to set budgets at the project or team level, flag unusual spending patterns, and tie cost spikes back to specific changes – such as a new feature launch or a shift in model usage.
It also handles reconciliation between usage and invoices, helping ensure that what companies are billed matches what they actually consumed, and that those costs are correctly assigned across the business.
A new layer in the AI stack
The launch reflects a broader shift in how companies are thinking about AI infrastructure. Much of the existing tooling around AI usage has focused on developers, with dashboards that track latency, performance, and prompt behavior.
While such observability tools are important, Ramp is angling for a layer that sits alongside them, aimed at finance and operations. The goal, ultimately, is to answer a different set of questions: not how a model performs, but what it costs and whether that cost is justified.
While that distinction will become more important as AI moves from experimentation into core business operations, Ramp is also making a case that better visibility could influence how businesses deploy AI in the first place. With clearer data on costs and returns, teams may adjust how they use models, which features they prioritize, or how they price their own products.
“The companies that build financial discipline around it now will make better decisions about where to invest, where to cut, and how to price their own products,” Atiyeh said. “The ones that don’t will be guessing.”
YOUTUBE.COM/THENEWSTACK
Tech moves fast, don’t miss an episode. Subscribe to our YouTube
channel to stream all our podcasts, interviews, demos, and more.
SUBSCRIBE
Group
Created with Sketch.
Paul is an experienced technology journalist covering some of the biggest stories from Europe and beyond, most recently at TechCrunch where he covered startups, enterprise, Big Tech, infrastructure, open source, AI, regulation, and more. Based in London, these days Paul…
Read more from Paul Sawers
