Navigating rising AI costs is getting tricky

Sponsored by

Ready to move past the Wild West of AI experimentation? Try this decision-making framework that PwC’s experts came up with. It outlines where to lead, lag, or exit on AI strategy. Get the details.

Costs associated with AI use are becoming dizzying for companies. Tokenmaxxing (equating higher AI usage with productivity, according to IT Brew), duplicative processes, and the sheer velocity with which AI has grown in the last few years has created what procurement software platform Tropic’s CFO Russell Lester called an “AI tax.”

“You’ve got software companies trying to transform into becoming AI businesses in varying degrees, whether it’s AI native or AI wrapper or some type of agentic approach,” Lester said. “Many of them are bolting on those capabilities in their pricing model, and others are being a little more savvy than that, but one way or the other, it is disrupting and showing up in software purchases and renewals.”

“The new world for a CFO to navigate is how to figure out how to budget for all of this, and how do you forecast for something that’s changing weekly?” Lester said.

In this new world, Lester said finance heads must be dogged in leading and tracking AI deployment across corporate divisions, not just on their teams.“CFOs need to lead the charge in standing up discipline to memorialize learnings and insights, and foundational layers of how we use AI at the corporate level,” he added.

Inconsistent pricing. Through working with CFO clients and as Tropic’s CFO, Lester said it has become easier and easier for companies to overconsume AI, given that tokens or credits are used differently across the industry.

“The definition of a token or a credit is not consistent across supplier[s], and as you’re using tools, it’s often opaque how much of it you’re consuming,” he added.

Consistency is key for a CFO; it’s hard to forecast, deploy capital wisely, or reduce risk when working with inconsistent figures.

“The combination of those is hard, but I think AI can both be a reason why it’s harder, and a catalyst for change to make it easier,” Lester said. “Because for the same reason that AI on the pricing front is making budgeting harder, AI itself can make the prediction of those things easier.”

Tool rationalization. To guide AI usage, Lester said to begin with questions like, “What are we hiring this particular investment to do? What are the outcomes it needs to achieve? What company priorities and goals does it tie into?”

This can help prevent blanket AI applications for all functions across a company, where every employee is “blowing through all their credits very quickly.”

Using the example of Claude Skills—instructions, scripts, and resources for domain-specific tasks—Lester said CFOs should think about creating a “skills mindset across the enterprise where there’s one skill for brand standards, and one skill for all the definitions that we use around data.”

“What’s our preference for how charts look? What are our key systems and how do they talk to each other? You don’t want each employee doing that on their own inefficiently, because then you’re reinventing the wheel a ton of times. So CFOs will have to not only figure out, ‘Are we buying the right tools, are we paying the right price, are we eradicating spoilage?’ But ‘Is all of this leading to stronger outcomes?’” Lester said.

Regardless of the AI deployment or application, he said the office of the CFO is “almost uniquely qualified” to meet the current AI moment because of the function’s ubiquity in an organization.

“No one else wakes up and thinks about the same things that a CFO does. A CFO is looking under every rock, talking to every team; they can see the data that’s confidential, they’re part of conversations other people are not,” Lester concluded.