What Enterprise Teams Are Actually Saying About AI Cost Management in 2026 | Yarken
What Enterprise Teams Are Actually Saying About AI Cost Management in 2026 | Yarken
At FinOps X 2026 in San Diego, we spent four days talking with FinOps practitioners, IT finance leaders, and technology executives from some of the world's largest enterprises.
What came out of those conversations wasn't just optimism about AI. There is a real sense of urgency to establish stronger governance, with greater visibility over AI spend.
This post captures what we heard directly from the teams living this problem; the gaps they're trying to fill, the tools they're evaluating, and the pressure building from leadership to prove that AI investment is delivering business value. We've also pulled in data from the FinOps Foundation's 2026 State of FinOps report because what we heard on the floor lines up almost exactly with what the data shows.
If you're a FinOps practitioner, an IT finance leader, or anyone currently trying to answer the question "what is our AI actually costing us?" this is where your peers are right now.
The FinOps Mission Just Changed. Here's What That Means.
Before getting into what we heard on the ground, it's worth noting what the FinOps Foundation announced this year. The Foundation updated its official mission from "Advancing the People who manage the Value of Cloud" to "Advancing the People who manage the Value of Technology."
One word swap, cloud to technology, but it reflects a fundamental shift in what FinOps teams are actually responsible for.
FinOps is no longer a cloud cost discipline. It is a technology value management discipline.
That means AI spend, SaaS, software licensing, on-premises infrastructure, and data center costs are all now in scope. The mandate has expanded and the tooling, skills, and frameworks most teams have built are still catching up. This is exactly what the conversations on the FinOps X floor felt like.
AI Spend Is Now Universal. Governance Isn't.
The State of FinOps 2026 puts a number on something practitioners already feel. In 2024, only 31% of respondents reported managing some form of AI costs. That jumped to 63% in 2025. In 2026, it reaches 98%.
AI is now a core operating cost. Full stop. But the conversations we had at FinOps X told a more complicated story. Almost every team we spoke with is managing AI costs in some form. Very few of them feel like they have it under control.
Organizations can see that their AI spend is growing. Hardly anyone can see why, who is driving it, or what value it is generating. Token counts are visible. Business outcomes are not. And the gap between those two things is where the real problem lives.
We heard from teams where AI spend had gone from a managed budget line to an exponential cost pressure in under twelve months. Thousands of developer licenses, model inference costs scaling faster than anyone planned for, GPU utilization that doesn't map to any existing reporting framework. In one case, a company was moving from a few hundred AI tool licenses to thousands, with no governance layer in place to understand what that investment was actually producing.
The pressure is coming from leadership. And most teams are not yet equipped to respond with the kind of structured, value-linked analysis that leadership actually needs.
Why Token Counting Isn't Enough
A significant portion of the market right now is solving an important, but narrow problem. At FinOps X, the dominant vendor conversations were about tokenomics: telemetry, token tracking, visibility into inference costs.
These are useful capabilities. But they answer the question "how much are we spending on tokens?" They do not answer "what is that spend producing?" The FinOps Foundation's own framing confirms this. The gap practitioners are struggling with is not just visibility into spend. It is connecting that spend to business value.
Many organizations are simultaneously being asked to self-fund AI investments through FinOps efficiency gains, which requires directly linking optimization work to strategic AI enablement.
The challenge is structural. AI pricing models based on tokens, inference requests, and GPU utilization don't map cleanly onto billing frameworks built for traditional infrastructure.
Token visibility and management are important aspects of effectively governing enterprise AI, but it has to go further; mapping shared costs, labor, hardware depreciation, and more to fully understand the trade-offs. Connecting spend to what the business is actually getting in return is critical for control of your organization’s AI economics.
The Siloed IT Spend Problem Is Getting Worse
One of the clearest patterns across our conversations was the persistence of siloed IT finance. Teams managing cloud costs separately from SaaS, separately from licensing, separately from AI, and separately from on-premises infrastructure. Each silo with its own tools, its own data, its own reporting cadence.
This was already a recognized problem before AI entered the equation. AI has made it significantly worse, because AI costs don't sit cleanly in any one of those buckets. A single AI workload can touch cloud compute, licensed model APIs, internal engineering time, and SaaS tooling simultaneously. If your reporting infrastructure is siloed, you have no way to see the true cost of that workload.
According to the 2026 State of FinOps report, 90% of respondents manage SaaS spend or plan to, up from 65% in 2025. 64% manage licensing, up by 15%. The scope is expanding. But expanding scope into existing silos just creates more complexity without creating more clarity.
Across our conversations at FinOps X, the picture was more fragmented than a single gap. Some teams had no platform at all, managing everything manually or through spreadsheets. Others had built homegrown solutions for cloud but had nothing covering the full scope of IT spend, and were not yet ready or mandated to go further. Some were on legacy platforms they did not like, found too expensive, and felt were not keeping pace with where their needs were going. And a growing number were in a holding pattern, waiting for the right moment to make a move, with AI costs now accelerating that timeline faster than they expected.
The common thread was not where they were. It was that the gap between what they had and what they needed was becoming harder to ignore.
What IT Finance Leaders Are Evaluating Right Now
Based on our conversations at FinOps X, here are some of the top use cases for teams actively evaluating value management platforms in 2026:
Unified cost visibility across all IT spend categories
Not just cloud. AI, SaaS, licensing, on-premises, and infrastructure costs in a single view. Teams currently operating on spreadsheets or siloed point tools are actively looking for consolidation. The driver is usually a leadership demand for one coherent number: "what are we spending on technology, and what are we getting for it?"
AI Economics and cost-to-value alignment
Multiple teams specifically named AI Economics, not token management, but the broader question of connecting AI investment to business outcomes, as the capability they are either looking for or actively building. Some are building internally. Many are evaluating whether a purpose-built platform is the faster path.
Replacing tools that weren't built for this era
A consistent theme across conversations was frustration with existing platforms around data ingestion flexibility, pricing, and the ability to handle FOCUS data formats and modern AI cost structures. Several teams described being mid-evaluation or mid-RFI, with active timelines in 2026.
Total Cost of Ownership for AI
AI TCO is emerging as a distinct use case. It goes beyond running cost and includes infrastructure allocation, labor, model depreciation, and the cost of shared model infrastructure that is difficult to attribute accurately. Teams responsible for enterprise-level AI programs want to produce a defensible TCO number, both for governance and for making the case to leadership that AI spend is justified.
Chargeback and showback for AI costs
As AI costs grow, allocation back to business units becomes politically necessary. Which teams are consuming how much model inference? Which products are driving GPU utilization? Answering those questions requires an allocation model that most current tools are not set up to handle.
The Organizational Reality: Small Teams, Big Scope
One thing that came through clearly in both the floor conversations and the 2026 report is the structural pressure on FinOps teams. FinOps teams stay small even as scope expands. The most common structure is centralized enablement at 60%, followed by a hub-and-spoke model at 21%.
Even in large environments, team size stays lean. Among organizations spending over $100M on cloud annually, the average FinOps team is roughly 8 to 10 practitioners plus a few contractors or service providers.
Small teams absorbing an expanded mandate. That is the operating reality for most of the practitioners we spoke with at FinOps X. They do want more headcount. But in the absence of it, the case they are making to leadership is one of two things: a platform that costs less than what they are already paying for legacy tooling, or a platform that surfaces savings fast enough to fund its own existence and justify the hire.
The goal is not to replace people. It is to give the people they have the tools to move faster and make better decisions. This shapes what a unified platform actually means in 2026. Point solutions add capability in isolation. What these teams need is a platform that brings everything together, reducing the time between "we have a question about AI spend" and having an answer in front of the CFO in under five minutes, without the manual effort that a report like that would typically require.
This is where AI-native platforms are changing the equation. The same organizations generating exponential AI demand are now enabling their FinOps teams to keep pace with it. Agentic workflows, automated cost allocation, and intelligent anomaly detection are compressing work that used to take days into minutes. Small teams are not just surviving the expanded mandate. With the right platform underneath them, they are meeting it.
What We Took Away From the Market
Two things stood out from the FinOps X experience that are worth naming directly for anyone building their strategy for the second half of 2026.
1. The market is validating the TBM + FinOps convergence story.
The FinOps Foundation's mission update is not a rebranding exercise. It reflects where practitioners have already been operating. Teams that have historically sat in FinOps are being pulled into TBM territory, chargeback, business value alignment, enterprise-wide cost allocation, because AI and SaaS sprawl have made those problems unavoidable. The disciplines are converging in practice. The platform needs to reflect that.
2. Peer credibility moves faster than any pitch.
Enterprise IT finance teams do not make decisions based on vendor pitches. They make them based on trust in other practitioners. At FinOps X, the moments that moved conversations forward were not polished demos. They were exchanges where someone respected in the room had already seen the platform and said so. A positive comment from someone respected in a live setting carries more weight than any vendor case study. At FinOps X, those moments happened organically and repeatedly. That is not something you can manufacture. It is something you earn by showing up consistently, delivering real value, and letting the community do what communities do.
Where This Is Going
The teams at FinOps X are not waiting for the market to settle. AI costs are already material, already growing, and already creating the kind of leadership pressure that accelerates buying decisions.
The question is not whether to build better AI cost governance. It is whether your current platform can support the speed, flexibility, and accuracy of analysis your leadership is going to demand in the next 12 months.
Most teams we spoke with already knew the answer. They just did not have the platform to change it.
Yarken is the only AI-native unified TBM + FinOps platform designed to help you manage value across your entire IT estate (across cloud, SaaS, infrastructure, etc.) including the exponential proliferation of AI technologies across the enterprise.
If you want to talk through your current setup, book a session with our team.