If you asked your finance team today how much each department is spending on AI, would they have an answer?
If not, you're not alone. Nearly half (46%) of CFOs in Deloitte’s Q2 2026 CFO Signals survey named cost uncertainty or lack of transparency as their top concern about AI. And it's easy to see why.
Many AI tools charge based on usage, and departments often use a combination of tools, each billed differently. So your team has to piece together invoices and usage data from different vendors just to calculate total AI spending.
Shared API keys add another layer of complexity. When several teams use the same key, their usage gets lumped together under one account, with no easy way to tell which team incurred which charges.
As a result, it’s tough to get a clear view of today’s AI spend, let alone plan for next quarter or identify where usage could be more efficient. AI spend management helps you get that visibility and act on it.
In this guide, we’ll explain:
It draws on insights from Vena’s CFO Melissa Howatson and CTO Hugh Cumming, shared in a recent Acterys and Vena webinar on managing AI consumption costs.
AI spend management is the process of tracking, forecasting and controlling your organization's spending on AI. These costs include anything from software subscriptions to API usage charges.
With traditional software, you could set a budget by multiplying the number of licenses by a fixed price. Some AI vendors share a similar model, charging a fixed fee per user, with plans that include a set number of AI credits. If you exceed that allowance, you’ll need to buy more credits, pay as you go, upgrade your plan, or wait until it resets.
Other AI vendors charge directly based on usage, so your bill can vary each month.
For AI services, such as the Claude API, consumption costs are calculated by token, particularly when teams connect directly to AI models to build their own tools.
A token is a small chunk of text such as a word, part of a word, or punctuation mark. You pay for both input tokens (what you send to the model, such as questions and documents) and output tokens (what the model writes back).
The more people and apps in your organization use AI, and the more text each request involves, the more tokens you use. Premium AI models, such as Claude Opus 4.6, generally cost more per token than smaller models such as Claude Haiku 4.5. That’s why the same task, for example summarizing a report, can cost different amounts depending on the model and how long the response is.
Understanding how token usage translates into costs (called AI tokenomics) helps you estimate future bills and find ways to get the same work done for less.
Managing AI spend helps you keep costs within budget as adoption grows. It also makes sure that spending delivers value for your business.
As more employees and teams use AI in their everyday work, what starts as a small expense can quickly become a significant business cost. Take Uber, for example. The company reportedly used up its full-year 2026 budget for AI coding tools in about four months, according to Fortune.
For finance leaders, those ballooning costs bring new responsibilities. Vena CFO Melissa Howatson explains it this way, “It’s a line item that went from someone else’s problem to our problem in Finance in the span of just 12 short months, without a budget, without an owner and without a forecast.”
But reviewing invoices after the money has been spent gives your finance team little opportunity to keep those costs within budget. That’s why your team needs to look ahead at what’s driving usage and how it could affect future spending.
“We need to get away from just saying, ‘What did we just spend?’ to asking, ‘What are we likely to spend? What are the drivers of consumption, and what’s the value we’re getting from the money that we’re spending?’” says Melissa.
Answering those questions helps you spot rising costs early, adjust your budget and assess whether the additional spending is delivering enough value.
The five-step process we’ve outlined below should help you track your AI spending, forecast future costs and establish who’s responsible for managing the budget.
You might already use your AI vendors’ dashboards to track usage and costs. But those alone won’t give you a complete picture of spending across your organization or help you see what you're likely to spend next.
“A monthly retrospective review on token consumption is insufficient,” says Vena CTO Hugh Cumming. Instead, he recommends tracking usage as it happens and feeding that data into your financial workflow, so you can start building forecasts from it.
To build that visibility, first list every AI tool your company pays for, including any that departments signed up for on their own. Note who uses each one, its pricing model and what it costs today. Then bring that information into a spreadsheet or planning tool, set monthly spending limits and compare them with actual costs.
That way, you’ll spot tools nearing or exceeding their limits early. Some AI vendors offer spending alerts, so turn those on to catch spikes between reviews.
If you're tracking AI spending manually in Excel, aim to update your figures at least weekly, or more often if usage is growing quickly. Waiting until the end of the month could leave you with unexpected costs.
If you'd rather not build your own spreadsheet from scratch, Vena offers a free AI spend planning template for Excel with a ready-made layout for this.
Next, make sure every cost can be traced back to the team behind it.
As Hugh puts it, “You have to work as a partnership between Finance and IT to manage and control which tools different parts of the organization have access to, and then create the pipelines to build these real-time views.”
Finance, IT and department leaders should agree on which AI tools teams can use and how the costs will be allocated across departments. IT can then put tracking in place so your finance team can see what each team is spending.
For AI tools accessed through APIs, work with IT to set up separate projects or workspaces for each team, where your vendors support this. Where possible, give each team its own API keys rather than having everyone share one.
AI vendors like OpenAI and Anthropic let you track API usage and costs by project or workspace. You can download reports from their dashboards and add the figures to your spending tracker. As usage grows, IT can use the vendors' usage and cost APIs to pull data into your tracking or reporting system automatically, for example once a day.
If teams shared accounts in the past, those bills may be impossible to split accurately. In that case, agree on a reasonable way to estimate each team's share, such as using the number of users in each department, and focus on precise tracking from there on.
Knowing why your AI bill is rising helps you find ways to reduce costs without stopping employees from getting value from the tools they use.
Start by comparing spending across teams, tools and AI models from one month to the next. If costs jump, look at what changed. Sometimes, the difference comes down to how employees use AI.
For example, Melissa recalls how a team at Vena discovered that asking AI to produce a presentation in a specific format cost about 10 times more than letting the model handle the layout. Once you know what kinds of requests use more tokens, you can decide when a particular format is necessary and when the AI-generated version is good enough.
Other reasons your AI costs might increase include:
Once you’ve delved into the likely drivers behind rising AI spend, share what you find with department leaders so their teams can make more informed choices and pick the right AI tools for each task.
Forecasting helps you make room for growing AI usage before those costs catch you by surprise.
Start with your current spending and usage data, then ask department leaders what they expect to change next quarter. Are they planning to roll out AI to more employees, automate additional tasks or switch to different models?
Then use that information to build three scenarios:
As you work through these scenarios, keep in mind that the impact on your budget depends on how each tool is priced.
Seat-based costs generally increase when you add users, while consumption-based costs depend on how much you use. Hybrid plans include a fixed allowance, with extra charges once you exceed it.
For example, if your engineering team expects to make 20% more AI requests next quarter, what would that mean for its budget? The answer depends on how its tools charge and which models handle the work. A 20% increase in requests won't always mean a 20% increase in spending.
You can use Vena's free AI spend tracker to model how spending may ramp up, slow down or shift through the year as adoption changes. Then bring these estimates into your financial plan and compare them with actual spending throughout the quarter.
If your plan lives in Excel, using built-in AI to analyze your finance data can help you spot where actuals are drifting from the forecast faster.
“Within our organization, we keep going back and re-forecasting what we’re going to spend on AI,” Hugh says. “It’s become part of how we do work every day. But it’s a material spend that we, like many other organizations, have had to make room for.” If usage grows faster than expected, you can update the forecast and decide whether to adjust the budget or find savings elsewhere.
Without a clear budget owner, AI costs can rise without anyone taking responsibility for keeping them in check. To avoid this, make each department responsible for its own AI budget, since they’ll have a better operational view of where AI helps them do their work better and which tools are worth keeping.
That’s how we’ve approached it within Vena. Melissa explains that the company initially paid for its first AI tool through the IT budget. As more teams requested tools, Finance and IT worked together to review the requests and decide which tools to approve. They then began allocating the costs to individual departments. Today, business units own their AI budgets, while Finance and IT help them track spending and decide when additional investment makes sense.
You can take a similar approach. Assign a budget owner, usually the department leader, to review costs and approve additional spending.
If you can't yet tell exactly how much each department spends on AI, start by showing teams their estimated costs without taking the money from their budgets. This is called “showback.” Once you can reliably track what each team spends, you can move to “chargeback,” where those costs come out of the department's budget.
Next, work with IT and team leads to set monthly spending limits based on your forecasts. Set up alerts so teams know when they're approaching those limits.
If a team reaches its limit, don’t automatically cut off access. Hugh sees that moment as a chance to talk with the team about what they’re trying to do and whether the extra spending is worth it. For example, if they need more AI usage for a project that’s saving employees several hours each week or helping close more sales, raising the limit makes more sense than restricting access.
The goal is to keep spending under control without discouraging employees from experimenting with AI or finding valuable ways to use it.
Your invoices show what AI costs, but not what it’s worth. “When you look at the invoice, it’d be very hard to decide how it was allocated, whether it was good spend or bad,” Hugh says.
To find out whether your AI spending is paying off, define what success looks like for each major use case. Record your starting point before introducing AI so you can measure what changes.
Some useful metrics include:
Then, compare the results with what you’re spending. For example, say an AI tool costs $2,000 a month and saves your finance team 100 hours of manual work. If an employee's average loaded cost is $60 an hour, that's $6,000 worth of time saved each month. So you're spending $2,000 on a tool that frees up $6,000 worth of employee time. That’s time your team can put toward other work.
Be sure to review the results with each budget owner at least quarterly.
If an AI tool is delivering strong returns, consider raising its spending limit or expanding access. If the returns are disappointing, check whether teams need training or could use a cheaper model before cutting back. And if the value still isn't clear, find out why before redirecting the budget elsewhere.
As more teams find new ways to use AI, spending will likely keep growing. What matters is whether you know where the money is going and have enough information to decide which investments are worth expanding.
With the right process, you can spot unexpected increases before they become budget overruns, make room for new AI projects and help teams get more value from the tools they use.
You don’t have to build everything at once. Start by listing your AI tools, what they cost and who’s using them. Vena's free AI spend tracker for Excel gives you a starting point for organizing those costs and forecasting future spending.
And if your team already works in Excel, see how Vena and Claude in Excel work together to help finance teams analyze data more easily. Or follow our step-by-step guide to using Claude in Excel with Vena.
Download Vena's free AI spend tracker.
AI token spend management is the process of tracking and controlling how much your company pays for tokens. It helps your finance team see which teams use the most tokens, what's driving those costs and how much to budget for future usage.
AI tokenomics focuses on understanding how the mechanics of AI consumption—such as the amount of text an AI model processes—affects its cost. For example, asking AI to summarize a short email typically costs less than asking it to analyze a lengthy report. The price also depends on which model you use and how much text it generates in response. Understanding these differences helps you estimate what AI tasks might cost.
List every AI tool your company pays for, who uses it and how it’s priced. Vena’s free AI spend tracker for Excel gives you a ready-made place to record each vendor, its pricing model and a monthly spending limit. Then compare those limits with the bills and usage reports you receive from each vendor. Where possible, give teams separate accounts so you can tell who’s responsible for each charge. Review spending regularly instead of waiting until the end of the month.
There's no standard amount. Your budget depends on how many employees use AI, what they use it for, how often they use it and how each tool charges. Start with what you're spending today, factor in any plans to expand usage and allow extra room for unexpected increases.
Look for ways to get the same results at a lower cost. For example, employees could use cheaper AI models for simple tasks, and teams could stop paying for tools that do the same job. Set spending limits and review requests for more usage rather than automatically restricting access.