The next AI budget request lands on the table, and the first question is a simple one. What will we get for it? The usual answers sound reasonable. Unit costs are coming down, productivity should go up, and the business expects to run more efficiently. Each one expresses hope, but none of them ties the money to an outcome someone owns.
In a webinar hosted by Cprime, Carollyn Gehrke, Digital Product Manager at Cprime, and Lauren Bateman, who leads Technology Business Management (TBM) Finance at FedEx, shared how they approach AI cost management. Their advice was to build evidence of value on purpose, well before anyone asks for it.
AI consumption is up 30%. Is that good news?
Gartner expects worldwide AI spending to reach $2.7 trillion in 2026, almost 50% more than the previous year. Closer to home, finance says AI spend is up 20%. The platform team reports consumption up 30%. Leadership wants to know if that’s good news, and neither number can say. If usage grew because people adopted the tool and saved thousands of hours, you have a strong story. If not, you’re holding cost nobody can defend. Part of the problem is that the answer sits across strategy, portfolio, finance, and technology teams.
From pilot to production, the questions get harder
Picture a customer service team piloting an AI assistant. Early on, it needs a real problem, a modest budget, and one person who decides whether the pilot continues. When the team wants to roll it out, the bar goes up. It now needs a value hypothesis, a baseline such as current resolution times, and someone to confirm the improvement is real. Once agents rely on it daily, the questions turn to run cost and whether usage still supports the business case.
Finance’s role shifts too, from setting spending limits on experiments to confirming that run costs match the value the business signed up for. TBM for AI helps most when it covers that whole journey.
Every token needs a business case
Platform teams usually track compute, storage, tokens, and inference well. What’s missing is context, such as which priority a workload supports, which service benefits, and who owns the business case. Without it, a token count only tells you how busy the system was.
Suppose the assistant’s costs jumped last quarter. If adoption grew and resolution times dropped, that may justify more investment. If adoption stayed flat, the same cost line tells a very different story.
That context usually exists, spread across costing, planning, FinOps, and strategic portfolio management tools. TBM pulls it together and allocates cost to the services that use it. As Cprime’s look at cloud economics in the age of AI explains, FinOps for AI spots the spike in inference costs, and TBM shows which business service drove it and whether the outcome justifies it.
One AI bill, three owners
A workable split gives technology the platform and its efficiency, gives finance transparency and controls, and leaves the business accountable for the outcome. On a shared platform, each business unit needs to see what it uses and what it gets back. A good check is whether every cost can be traced to someone who answers for it, which gets easier once finance and IT are aligned through TBM.
Governance should grow with the investment. Keep early experiments short, capped, and owned by a named person, and add visibility as a service reaches more users or handles more sensitive data and critical work.
Measure AI ROI one decision at a time
You don’t need a complete AI cost model to start. Pick a real decision, like expanding a pilot or renewing a platform contract. Define success, baseline the cost and consumption behind it, and link that baseline to the strategy, owner, and funding source. Then agree on who reviews it, how often, and what results mean you continue, adjust, or stop.
At review time, compare results to the baseline, check unit economics like cost per transaction, and be honest about whether a pilot has failed or simply hasn’t had a fair run. The evidence only has to be good enough for the next decision.
Where Cprime fits in
Cprime, a TBM Council member, helps finance and technology leaders bring TBM, FinOps, and strategic portfolio management into one operating model. That work includes allocation models for shared platform costs and review cadences your teams can run themselves. Cprime’s Enterprise AI TBM Value Lifecycle Framework, covered in the webinar, follows each AI investment from pilot to production.
Four questions to answer before your next funding review
The next funding conversation will hinge on four things. What are you buying, for whom, why now, and what happens next? Without those answers, confidence in the next AI investment takes the hit.
Start with the AI decision on your calendar next quarter. Name the person who’ll make it, and work out the minimum evidence they need. To see the full framework, watch the webinar on demand.
Ready to prove the value of your next AI investment?
Cprime helps finance and technology leaders connect TBM, FinOps, and strategic portfolio management so every AI dollar traces to an owner and an outcome. Talk to our experts about building the evidence your next funding review will ask for.
Frequently asked questions (FAQs)
Is rising AI consumption a good sign?
Not on its own. A 30 percent rise in consumption or a 20 percent rise in AI spend cannot say whether the money is working. If usage grew because people adopted the tool and saved thousands of hours, that is a strong story. If not, it is cost nobody can defend.
How do AI cost questions change from pilot to production?
A pilot needs a real problem, a modest budget, and one person who decides whether it continues. A rollout needs a value hypothesis, a baseline such as current resolution times, and someone to confirm the improvement. In production, the questions turn to run cost and whether usage still supports the business case.
Why do AI token and inference costs need business context?
Platform teams track compute, storage, tokens, and inference well, but without context a token count only shows how busy the system was. Business context shows which priority a workload supports, which service benefits, and who owns the business case, so a cost spike can be judged against adoption and outcomes.
How do TBM and FinOps work together on AI costs?
FinOps for AI spots the spike in inference costs. TBM pulls together context spread across costing, planning, FinOps, and strategic portfolio management tools, allocates cost to the services that use it, and shows which business service drove the spike and whether the outcome justifies it.
Who should own AI costs?
Split ownership three ways. Technology owns the platform and its efficiency, finance owns transparency and controls, and the business is accountable for the outcome. On a shared platform, each business unit should see what it uses and what it gets back, and every cost should trace to someone who answers for it.
How do you measure AI ROI?
Start with one real decision, such as expanding a pilot or renewing a platform contract. Define success, baseline the cost and consumption behind it, and link it to the strategy, owner, and funding source. At review, compare results to the baseline and check unit economics like cost per transaction.
What questions should you answer before an AI funding review?
Four: What are you buying, for whom, why now, and what happens next? Without those answers, confidence in the next AI investment takes the hit. Start with the AI decision on your calendar next quarter, name the person who will make it, and work out the minimum evidence they need.
Ready to prove the value of your next AI investment?
Cprime helps finance and technology leaders connect TBM, FinOps, and strategic portfolio management, so every AI dollar traces to an owner and an outcome. Talk to our experts about building the evidence your next funding review will ask for.
