The CTO's AI Spend Conversation With the CFO Has Changed. Here's the New Script.

The old AI spend conversation was about licenses, adoption, and vendor productivity claims. Now that AI costs are spread across seats, tokens, agents, and usage-based charges, that’s no longer enough. The CFO needs to know what the organization spent, where the money went, what changed as a result, and what the next budget period is likely to cost. That changes what the CTO needs to bring to the conversation. Key Takeaways AI cost management is now mainstream. The 2026 State of FinOps report found that 98% of respondents manage AI spend, up from 31% two years earlier. The 2026 FinOps Framework added Executive Strategy Alignment as a capability, reflecting a shift from technology cost reporting toward connecting investment with business priorities, multi-year planning, and executive decisions. A credible CTO-CFO conversation needs four connected views: spend by team and tool, engineering outcomes associated with that spend, a forward-looking forecast, and controls for managing unexpected costs. What the CFO Is Actually Asking Once AI becomes a meaningful operating expense, “Are people using it?” is not enough. Finance needs answers to questions such as: Which teams and tools are driving the cost? Is higher spending associated with better delivery or quality? What will this cost next quarter? Where could spend exceed budget? Which costs still cannot be attributed to an owner or use case? The 2026 FinOps Framework reflects that shift, positioning technology-value management as an executive decision-support function rather than simply a cost-reporting exercise. For CTOs, that means the conversation works better when cost and engineering data arrive together. The Four Data Points That Change the Conversation 1. Spend by Tool, Team, and Use Case Start with where the money is going. A total engineering AI bill is useful for accounting, but it doesn’t tell a CFO what’s driving the change. Break spend down far enough to show which tools, teams, agents, and use cases account for it. Larridin’s Token Spend & Insights connects spend to teams, agents, projects, workflows, models, and use cases. It also surfaces unattributed spend so finance can see which costs still lack a clear owner. That gives the CTO a much better starting point than a stack of unrelated invoices. 2. Engineering Outcomes Beside the Spend The next question is what the organization got for the money. That doesn’t mean converting every engineering metric directly into dollars. PR cycle time, code durability, rework, delivery frequency, and incident rates are operating signals. They’re useful in the finance conversation when leaders can show how they changed alongside AI adoption and spending. Our AI coding ROI guidance recommends framing time savings as capacity unlocked rather than automatically treating saved developer time as payroll savings. That distinction makes the financial case more defensible. The useful statement isn’t “PR cycle time fell 35%, therefore we saved $1.4 million.” A better statement is: “This team’s AI spend increased while PR cycle time fell, quality remained stable, and measurable engineering capacity increased.” 3. A Forecast, Not Just the Current Bill The CFO also needs to know what happens next. AI costs can change as adoption deepens, agentic usage expands, teams switch models, or token consumption rises. A backward-looking invoice can’t answer whether the current budget will hold. Token Spend & Insights projects spend by team and flags budgets at risk before the quarter closes. That turns the conversation from “here is what we spent” into “here is where we are headed and where we may need to adjust.” The forecast doesn’t need to pretend AI spend is perfectly predictable. A range with clear assumptions is more useful than a precise number built on weak assumptions. 4. Controls and Ownership The final question is what happens when spending moves outside the expected range. Controls can include budget alerts, ownership for agents and use cases, review of unattributed spend, dormant-license checks, and thresholds for investigating unusual consumption. Larridin’s current spend view, for example, flags projected overages, unattributed spend, orphaned agents, and dormant seats. This is where the CTO can show that the organization isn’t just tracking AI costs after the fact. Someone owns the spend, exceptions are visible, and there is a process for deciding when to investigate or intervene. The New CTO-CFO Script The conversation can be surprisingly simple when the underlying data is connected: Here’s what we spent, broken down by the teams, tools, and use cases driving it. Here’s what changed in delivery, quality, and engineering capacity alongside that investment. Here’s what we expect to spend next, including the assumptions and major sources of uncertainty. Here are the controls and owners in place if consumption moves outside that range. That’s a much stronger conversation than “we bought X licenses and the vendor says developers are faster.” Frequently Asked Questions What if we don’t have attribution data before the next CFO meeting? Start with the best defensible view you have. Pull actual spend from invoices and provider data, identify what can already be assigned to teams or tools, and clearly label what cannot. If you estimate team spend from headcount, label it as an estimate rather than measured spend. Show the current gap and explain how you plan to improve attribution. What if the CFO wants to cut AI costs across the board? Put cost and outcome data side by side before deciding. A team with high spend and strong delivery gains may warrant a different approach than one spending about the same with little measurable improvement. Across-the-board cuts are simple, but they can cut productive investment along with waste. How often should the CTO and CFO review AI spend? How often you review AI spend depends on how quickly it changes. Monthly reporting with a deeper quarterly review is a reasonable starting point. Teams with highly variable agent or API costs may need to check spend more often, even if executive reporting stays monthly. How should CTOs translate engineering productivity into financial terms? Be careful about turning engineering gains directly into dollars. Faster delivery can create value through more capacity, earlier releases, less rework, lower outside spend, or fewer incidents. But saved engineering time does not automatically mean cash savings. Our AI coding ROI guidance recommends accounting for rework and framing recovered time as productive capacity rather than implying that every saved hour reduces payroll. Build the Data Behind the Conversation The CTO-CFO conversation gets easier when spend, ownership, forecasts, and engineering outcomes don’t live in separate systems. Larridin’s Token Spend & Insights provides the cost and attribution view, while developer productivity measurement connects AI activity with engineering outcomes. Together, they give leaders a clearer picture of what AI costs and what that investment is producing. Book a discovery call to build a clearer view of AI engineering spend and value. AI Token Spend Attribution: What CFOs Need to See How to Measure AI Coding Tool ROI for Engineering Leaders AI Token Spend & Insights Developer Productivity for CTO