The Token Bill Is Due. Treat Them for What They Are: Capital.

Over the last two years, most companies approached AI the same way they approach any emerging technology: experimenting broadly, learning quickly and giving teams room to figure out where the value is. That was the right approach because you don’t learn much about a new technology by over-governing it on day one. But the conversation is changing.

As AI usage expands across the business, what began as experimentation is becoming part of the cost structure. We’re already seeing SaaS companies introduce token-based pricing models and AI consumption bundles, making AI costs increasingly tied to usage. As spending becomes material, finance leaders need to think differently about ROI, accountability and resource allocation. 

The Shift From Adoption to Allocation

If AI is helping teams move faster, improve customer outcomes, make better decisions, or operate more efficiently, it needs a business case and justification like any capital. In some cases, it may be one of the highest-return investments a company can make, but there may also be cases where the return is not there.

If usage grows simply because the tools are available, costs can scale faster than value. That’s where finance must play its role. We can help the business allocate resources intentionally while still giving teams room to experiment and learn.

We’re working through that ourselves at Zuora. We want our teams to use AI and experiment with new ways of working. At the same time, as usage grows, we need visibility into consumption, costs, and outcomes. Balancing those priorities is quickly becoming a top priority for companies.

Three Questions Finance Should Be Asking
As AI becomes part of the cost structure, finance leaders need to get comfortable asking a few simple questions.

1. 1. What return are we actually getting from this usage? (Adoption doesn’t count) Adoption isn’t a return metric. Finance needs to understand whether AI spend is reducing cycle times, improving productivity, strengthening customer outcomes, improving decision quality, or creating another measurable business result. AI spend needs to either grow the top or bottom line. At Zuora, we’ve seen AI help automate portions of collections and cash application workflows, reducing manual effort, improving visibility, and removing operational bottlenecks that previously depended on individual team members.
   2. Where does premium usage actually matter? Not every workflow requires the most expensive model. At Zuora, we encourage employees to match the tool to the task, using lower-cost models for everyday work like summarization, research and knowledge look-up and reserving premium models for more complex or business-critical use cases. The companies that manage AI economics well match investment levels to business impact. Higher-cost resources should be used where the additional capability creates measurable value.
   3. What constraints help people make better decisions? Most leaders dislike constraints, but scarcity often improves allocation. We’ve introduced AI credit limits, starting with a soft-cap period to help employees understand their usage patterns and build awareness before stricter controls take effect.

The purpose is to create visibility into consumption patterns and the business value that is being generated. When teams understand that usage has a cost and resources aren’t unlimited, conversations become more thoughtful. People get more specific about use cases, compare alternatives more carefully, and think harder about where AI creates meaningful value.

The CFO’s Role Is Expanding Again

As AI moves from experimentation into operations, the CFO’s role extends beyond approving budgets and tracking spend. 

Finance must bridge the gap between the board’s and investors’ focus on ROI while allowing business leaders to keep experimenting. At the same time, technology will continue to evolve. Models will become cheaper, new vendors will emerge, and economics will change. The goal is to build a policy that can evolve with changes in the environment. We need to continuously evaluate whether the value being created still justifies the resources being consumed.

Finance leaders don’t need to become experts on every model, but we do need enough understanding to ask better questions. What drives cost? What drives variability? Which use cases are proving value?

The companies that get this right won’t be the ones that spend the least on AI. They’ll be the ones that build the best operating model around experimentation, accountability, and ROI. As AI becomes embedded in day-to-day operations, finance needs to apply the same rigor to these investments that applies to every other significant investment decision. 

What I’m Reading 

Monetization on Fury Road: AI Credits as Shock Absorbers

This issue focuses on a simple reality: AI consumption now has real financial consequences. In this article, Tien Tzuo explores why AI credits are emerging as a way to manage variable costs, unpredictable usage, and evolving monetization models. It’s a useful companion read for finance leaders thinking through the economics behind AI.

Supercharging Finance Productivity: Zuora AI Expands with New Agents

As finance teams look beyond AI experimentation and toward measurable outcomes, the question becomes where AI can create the most value. Shakir Karim shares how organizations are using Zuora AI across quote-to-cash to automate manual work, accelerate analysis, and reduce operational friction. A practical example of how AI delivers results when it’s embedded directly into the workflows teams use every day.