98% of FinOps teams are now managing AI costs. Two years ago it was 31%. That's not gradual adoption, that's a mandate that arrived overnight.
The spend was always going to come. What didn't come with it was the framework to explain it. There was no procurement process built in advance, no allocation model, no governance layer waiting on the other side. AI tools got approved, usage scaled, and by the time anyone thought to ask what it was costing the business, the invoices were already months deep with no structure to explain them. Most teams found themselves retrofitting governance onto something that had already grown past the point where retrofitting was comfortable.
And that's where it gets complicated. Because the question leadership is now asking isn't "how much are we spending on AI?" Token counts can get you close enough to answer that. The harder question, the one that's starting to land on FinOps desks more and more, is "what is it actually producing?"
That's a fundamentally different problem. A single AI workload can touch cloud compute, licensed model APIs, internal engineering time, and SaaS tooling all at once. If your reporting infrastructure was built for traditional infrastructure, there's a good chance you have no reliable way to see the true cost of that workload, let alone trace it back to a business outcome anyone upstairs would find credible.
Most teams right now are reasonably well equipped to report the number. Where things get harder is defending it, contextualising it, and connecting it to the value conversation leadership is increasingly expecting to have.
Yarken's AI Economics capability was built specifically for this gap. Not token tracking, but connecting AI spend to business value across every cost dimension it touches, because visibility without context isn't really governance. It's just a bigger dashboard.