There is a cost problem growing inside most businesses right now, and most finance teams cannot see it clearly. AI spending, meaning the fees companies pay to services like OpenAI, Anthropic, and Google every time their software or staff uses those tools, has become one of the fastest-moving line items in corporate budgets. The bills do not look like normal invoices. They come in from multiple providers, reference technical codes, and give no indication of which team, project, or person drove the cost.
Ramp, the New York-based financial software company, launched a product this week that directly addresses this. It pulls AI cost data from providers including OpenAI, Anthropic, Google's Gemini, and Cursor into one place, organizes it by team, person, project, or individual access key, sends weekly briefings with spending trends, and alerts someone in finance before a budget limit is hit rather than after. The product is free to start, with no requirement to be an existing Ramp customer.
The numbers behind the problem are striking. AI spending across Ramp's customer base grew over 13 times between January 2025 and mid-2026. That is not a percentage increase. That is 13 times the original amount. Token prices, meaning the per-use cost of running AI, have actually fallen sharply over the same period. But usage has grown far faster than prices have dropped, so the total bill keeps climbing.
The reason costs spiral is structural. Traditional software has a fixed monthly price per user, which makes budgeting simple. AI tools charge based on how much they are actually used: every question asked, every document processed, every automated task completed adds to the bill. An employee using an AI coding assistant more than expected, or a software system looping through a task incorrectly, can run up thousands of dollars before anyone notices. Uber burned through its full 2026 AI budget in just four months after rolling out one AI coding tool widely across its engineering team.
Ramp's data from over 70,000 businesses shows that the median company spends around $2,200 per month on AI access fees, while the average is closer to $140,000 because a small number of heavy users drive enormous bills. The gap between those two figures tells you how uneven and unpredictable this spending is.
The practical value Ramp is selling is visibility first, control second. One of its customers, AngelList, discovered through Ramp's weekly briefing that it had been spending $10,000 a month unnecessarily because of a technical setting that no one in finance had known to look for. The engineering team fixed it the same day. Ramp says it finds potential savings worth roughly 12% of the average business's monthly AI bill.
One in three businesses using the tool, Ramp reports, found they were using expensive AI models for tasks that cheaper ones handle equally well. This matters because the price difference between the most powerful AI models and the mid-range ones can be 8 to 10 times per unit of use. Routing simple tasks to simpler models, and reserving the expensive ones for genuinely complex work, is where most of the savings come from.
Ramp is a well-funded company with a clear interest in being the place where businesses manage all their spending, AI included. The product is free as a starting point, and its more advanced controls sit behind a paid tier. That said, the problem it is solving is real and growing, and the basic version is worth knowing about. If your business is using any AI tools and your finance team cannot currently tell you what they cost by team or project, this fills that gap at no immediate cost.