- Why AI token costs are harder to manage than other software expenses
- The two layers of AI cost control
- What drives unexpected AI token cost increases
- How finance teams can track AI API costs by team and model
- How to set AI budgets that hold up month to month
- How to reduce AI token spend: Tactics by layer
- How Ramp supports AI token spend management
AI token spend grew 572% year-over-year from June 2025 to June 2026. Finance teams are catching up to what engineering already knows: AI isn't a line item you set once and forget. It's a recurring, fast-moving cost that behaves more like infrastructure than SaaS, and it needs the same oversight.
Controlling them takes action at two layers: in the engineering stack and in the finance stack.
Why AI token costs are harder to manage than other software expenses
Most software costs are predictable. SaaS is seat-based, cloud storage is capacity-based, and both follow patterns your finance team can budget against. AI token costs don't behave that way.
Token spend varies by model, by prompt length, by the number of calls your applications make, and by which teams are running which workloads. Month-to-month swings are common: the typical (median) business's AI spend swings by about 58% month to month. 61% of businesses average swings of 40% or more.
The median business uses 2 distinct AI vendors. The average is 2.8, spread across Anthropic, OpenAI, Google, and others. Without a unified view, you're reading multiple invoices without the context to know what changed or who drove it.
That's the core management problem. The spend is real: 8.4% of companies now exceed $10,000 per month, the level where AI becomes a formal budget line. But the visibility to govern it sits in systems finance teams don't control.
The two layers of AI cost control
Reducing AI token costs requires action at two distinct levels:
| Layer | Who owns it | What it controls |
|---|---|---|
| Infrastructure | Engineering / DevOps | Prompt design, model selection, caching, batching, request routing |
| Finance | Finance / Controllers | Budgets, cost allocation, chargeback, spend limits, anomaly detection |
Most cost reduction guides focus entirely on the infrastructure layer because that's where tokens are generated. But the structures that make cost reduction stick across teams live in the finance layer.
What drives unexpected AI token cost increases
AI token cost surprises follow predictable patterns: model drift to premium tiers, long-context inflation, multi-model sprawl, and missing team-level visibility. These drivers of unexpected AI token cost increases compound because no single team owns visibility across the stack—the invoice arrives only after the spend has already happened.
Model drift upward. Teams start with a budget model for internal tooling and quietly upgrade to a premium model for production use without a corresponding budget increase. Premium models represented 45.8% of tokens used in April 2026 but 55.9% of costs—the gap between token share and cost share is where budget overruns hide.
Long-context inflation. Nearly one in five Anthropic tokens processed are now long-context. Applications that pass large documents, full conversation histories, or rich system prompts in every request consume tokens at a rate that's hard to anticipate from initial testing.
Multi-model sprawl. Companies using 26 or more models averaged $26,562 per month. Without a clear policy on which models teams can use for which cases, they add models incrementally, and spend accumulates across contracts you didn't know you had.
How finance teams can track AI API costs by team and model
Tracking AI costs at the team and model level means closing the gap between what your AI providers report and what your finance system understands.
Your providers bill by token consumption. Your finance system needs to see cost by business unit, project, cost center, and model. That level of detail lets you hold teams accountable and spot where optimization is warranted.
A basic tracking framework covers four dimensions:
| Dimension | What to track | Why it matters |
|---|---|---|
| Model | Cost per model (Opus, Sonnet, GPT-5, Gemini, etc.) | Identifies model drift and premium model overuse |
| Team | Cost attributed to each business unit or project | Creates accountability and enables chargeback |
| Provider | Spend by Anthropic, OpenAI, Google, etc. | Surfaces contract concentration and redundancy |
| Type | COGS vs. OpEx classification | Separates customer-facing AI from internal tooling for accurate margin reporting |
The cost of goods sold (COGS) vs. operating expenses (OpEx) distinction matters more than it sounds. AI tokens that directly serve customers—a product feature, a support chatbot, or a code generation tool in your product—belong in COGS. Tokens consumed by internal tools belong in OpEx.
Mixing them distorts gross margin and makes it harder to model AI unit economics as usage scales.
How to set AI budgets that hold up month to month
If you set a single AI budget at the company level, it's likely to fail. AI costs aren't generated at the company level—they're generated by specific teams running specific workloads.
A more durable approach: budget by team and model category, not just in aggregate.
Start with a baseline. Pull the last three months of AI spend by provider and map it to the teams that generated it. Your own actuals are the right starting point. Industry medians are useful benchmarks, but your internal pattern is what matters for forecasting.
Set per-team spend targets. Give each team a monthly AI budget based on their current spend and planned workloads. This doesn't have to be a hard cap initially. A soft target with visibility is enough to change behavior.
Classify new workloads before they scale. Before a new AI feature or application goes to production, require a cost model: estimated tokens per request, expected call volume, and projected monthly spend at three growth scenarios. This is standard practice for cloud infrastructure. AI tokens warrant the same rigor too.
Review monthly, not quarterly. AI spend compounds fast. Month-to-month swings of 40% or more mean quarterly reviews catch problems too late to prevent budget overruns.
How to reduce AI token spend: Tactics by layer
What the data shows about AI cost efficiency
The gap between high- and low-cost AI teams isn't about which tactics they know. Ramp's data across thousands of businesses makes the cost difference concrete.
Prompt caching. If your applications send the same system prompt or document context repeatedly, caching stores that input and charges a fraction of the normal rate on subsequent calls. According to Anthropic, cache reads cost 90% less than standard input tokens, a substantial reduction for high-volume, repetitive workloads.
Model sprawl cost. Companies using four to ten AI models spend $28 per employee per month at the median. Companies using 26 or more spend $442—a 15x gap that has nothing to do with output quality and everything to do with ungoverned adoption.
Batch processing. Many AI providers offer reduced pricing for asynchronous batch requests—tasks where a seconds-fast response isn't required. Internal reporting, document processing, and bulk analysis are common candidates. Anthropic's Message Batches API reduces costs by 50% for async workloads.
About 74.5% of businesses connected to Anthropic's API have enabled prompt caching, compared with 50.8% for OpenAI. The share that haven't are still paying full input token prices.
At the finance layer
These tactics create the structures that make infrastructure optimizations stick and catch spending problems before they compound.
Spend limits by team. A monthly limit for each team's AI budget creates a forcing function: teams have to choose which workloads to prioritize and which to optimize. Start with soft limits (alerts, not hard stops) and move to enforcement as your tracking matures.
Chargeback to cost centers. When AI spend is allocated back to the teams that generated it, those teams have a direct incentive to reduce waste. This is the same mechanism that made cloud cost management effective in engineering organizations, and it works for AI spend.
Anomaly alerting. A 40% month-over-month spike is hard to catch reviewing invoices manually. Automated flagging of spend that deviates materially from a team's baseline gives finance the early warning to intervene before an overrun becomes a budget crisis.
How Ramp supports AI token spend management
Ramp's Token Spend Management gives you visibility into AI spend across Anthropic, OpenAI, and other providers, consolidated in one place and broken down to the model and team level.
You can see what's driving costs, distinguish COGS from OpEx, and get automatic flagging of anomalies and savings opportunities. You don't need engineering teams to export data or build custom reports.
Token Spend Management is now available to Ramp customers.
FAQs
You can reduce AI token spend by acting at two layers. At the infrastructure layer, prompt caching, model matching, and batch processing reduce tokens consumed. At the finance layer, team-level budgets, chargeback, and anomaly alerting create the oversight layer that makes optimization stick. The most effective programs address both.
You can track AI API costs by mapping provider invoices (which bill by token) to internal cost centers by team, project, and model. This requires a tool that consolidates spend across Anthropic, OpenAI, and other providers with COGS vs. OpEx classification and team-level attribution built in.
The most common causes are model drift upward (teams upgrading to premium models without a budget change), long-context inflation (applications that send large inputs in every request), multi-model sprawl (more models without a clear approval policy), and missing team-level visibility that prevents accountability before spending compounds.
Budget by team and model category, not just at the company level. Start with a three-month baseline, set per-team targets, require cost models for new workloads before they scale, and review monthly. Quarterly reviews miss the 40%-or-more swings that are common in AI spend.
“Most banks treat the back office as a cost to keep down. We treat ours as a return to compound, which is why we run it on Ramp. Now we put our clients on Ramp, too.”
Patrick Gaughen
President & COO, Hingham Institution for Savings
“Browserbase builds infrastructure so AI agents can do real work. Ramp is doing the same for finance. It’s not another tool. It’s a system purpose-built for AI-driven finance, and that’s why we chose Ramp as our financial operating system from day one.”
Paul Klein IV
Founder & CEO, Browserbase
“We used to pay up to $20k a year for our AP platform. With Ramp, we’re earning back well over that amount. That's money that belongs to the mission now, not to the back-office software.”
Heidi Coffer
Chief Financial Officer, Boys & Girls Clubs of San Francisco
“The tricky thing about corporate travel policy is timing. We didn't need a stricter policy. We needed the policy to show up earlier. With Ramp Travel, it finally does.”
Keith Frantz
Director of Enterprise Risk Management, Prosper
“We're accountable to our funders, our partners, and the families we serve. That accountability starts with how we manage every dollar. Ramp makes it easy for our team to spend wisely, track in real time, and keep overhead low so more resources reach the families navigating infertility.”
Rachel Fruchtman
CFO, Jewish Fertility Foundation
“Each member of our team has an outsized impact due to our focus on using high-leverage tools like Ramp.”
Lauren Feeney
Controller, Perplexity
“With Ramp, we haven’t had to add accounting headcount to keep up with growth. The biggest takeaway is that instead of hiring our way through it, we fixed the workflow so we can keep supporting the organization as we scale.”
Melissa M.
VP of Accounting at Brandt Information Services
“In the public sector, every hour and every dollar belongs to the taxpayer. We can't afford to waste either. Ramp ensures we don't.”
Carly Ching
Finance Specialist, City of Ketchum
