Token Economics: The component of the FinOps Foundation model that no one has built yet
The price of AI has fallen significantly over the past year. An analysis of 2.4 billion enterprise calls to AI models shows a 67% decline in the average price per million tokens, from $18.40 to $6.07, from the first quarter of 2025 to the first quarter of 2026 (Optimum Partners, 2026). Nevertheless, 93% of enterprises report exceeding their AI budget (McKinsey, July 2026), and the FinOps Foundation's own 2026 report finds that 73% of enterprises have experienced AI costs that exceed original projections.
This is not a contradiction. The FinOps Foundation itself articulates the fundamental relationship like this: cost equals price times quantity, and you manage it by either reducing the price or reducing the quantity. The price per unit is falling. However, consumption per task is growing faster than the price is falling, and Goldman Sachs has projected that total token consumption could grow approximately 24-fold towards 2030. For a CFO or CIO, the question is therefore not whether AI is getting cheaper. It is. The question is whether anyone in the organization can explain why the bill is rising anyway, and whether there is anyone who can actually do something about it.
It is no longer just an IT bill
The decisions driving AI consumption are rarely made in the finance department. They are made when a team selects a model, designs a workflow, or allows a pilot project to go into production without recalculating the budget. The consequence only lands as a line item on the invoice months later, often without any clear connection to the decision that initiated it.
The FinOps Foundation's own framework recognizes how broadly responsibility is distributed. The framework defines specific personas for precisely this, including Finance, Engineering, Leadership, and Product, all of whom are involved in decisions that drive consumption. This is the very point: consumption is not owned in a single place. It is generated across roles that historically have not had a shared process for it. Indeed, for the first time, AI has become its own formal Scope in the FinOps Foundation's 2026 framework, alongside public cloud, SaaS, data centers, and licensing, precisely because it requires its own systematic approach.
This is one of the reasons why 49% of organizations have already postponed or scaled back AI initiatives due to costs (KPMG, Global AI Pulse). It is not because AI does not deliver value. It is because no one could explain where the cost originated, or how it related to the value the initiative actually created. It is a management decision made too late because visibility arrived too late.
The three questions a financial statement cannot answer
The FinOps Foundation's own framework is structured around four domains: Understand Usage & Cost, Quantify Business Value, Optimize Usage & Cost, and Manage the FinOps Practice. Translated into AI consumption, this practically becomes three questions an organization should be able to answer precisely.
Where is consumption being generated?
In the FinOps Foundation framework, this belongs to the Understand Usage & Cost domain. Not just which vendor, but which application, which team, and which workflow. Without that level of detail, an increase in AI costs is just a number no one can act upon.
How is it being consumed?
This belongs to the Optimize Usage & Cost domain, and to the Governance, Policy & Risk capability under Manage the FinOps Practice. A workflow that looks inexpensive on paper can in practice cost fifteen to twenty times more because it retrieves more context, uses a model that reasons in multiple steps behind the scenes, or initiates an agent to chain multiple actions together. This is not faulty budgeting. It is that the original estimate never accounted for how consumption actually grows when something is put into serious production.
What business value does it create?
This belongs to the Quantify Business Value domain, and this is also precisely where the FinOps Foundation, in their 2026 framework, has added a brand-new capability: Executive Strategy Alignment, which explicitly connects technology spend to business strategy and helps leadership weigh investments and make strategic decisions. This indicates how central the question has become. At the same time, it is the question that almost no enterprises can answer precisely today, because it requires connecting concrete consumption to a concrete business process and its outcome.
💡 A dashboard can answer the first question. It can rarely answer the second, and almost never the third. That is the difference between measuring a bill and being able to manage it. |
Worth keeping an eye on: Beside the FinOps Foundation, the Linux Foundation has announced an intent to establish a Tokenomics Foundation to build common, vendor-neutral standards for AI consumption, much like the FOCUS specification did for cloud. No published standards or finalized governance structure exist yet, only a declaration of intent from the FinOps X conference in 2026. This is a sign of how quickly the field is maturing, and something we are monitoring, but for now, the FinOps Foundation's own established framework remains the solid reference.
The ownership problem behind the bill
Behind the lack of answers typically lies not a technical limitation, but an ownership problem. Only 21% of enterprises currently have a mature governance model for autonomous AI agents, even though 23% already report at least moderate use of them (Deloitte, State of AI in the Enterprise 2026). A major study from SAP LeanIX finds that 98% of enterprises have already deployed or plan to deploy AI agents, but less than half have real visibility into how many agents are actually running. And at the pure cost level, Flexera’s 2026 State of ITAM Report finds that only 31% of enterprises have accurate visibility into their AI software spend, while 59% experience that their wasted AI spend has increased year-over-year.
This is also precisely what the FinOps Foundation’s own Governance, Policy & Risk capability points to: establish a governance model with clearly defined roles, and assign named owners to monitor costs, budget, and optimize consumption. Without this, a token budget will never be enforced, and ownership without financial context will never be prioritized because the owner cannot see what is actually at stake.
The architectural foundation that the framework itself assumes
There is a detail worth being clear about. The FinOps Foundation’s domains describe what an organization needs to achieve, not how to practically connect consumption to an application, an application to a business process, and a process to a named owner. This connection, from system to process to owner, is not a FinOps discipline. It is an Enterprise Architecture and Business Architecture exercise, and it is the one that actually determines whether the Understand Usage & Cost and Quantify Business Value domains become anything more than an intention on a slide.
This is also where Clariox's core expertise lies. We are not a FinOps vendor, and we do not build your finance or governance processes for you. We are an Enterprise Architecture and Business Architecture consultancy and Ardoq partner—the only Ardoq-focused partner in Denmark—and our role is to deliver the architectural foundation that your finance, IT, and business leadership need to actually act on the questions the framework raises.
What a cost dashboard can show, and what an architectural view requires
Question | Can a cost dashboard show this? | What does it require? |
|---|---|---|
How much did we spend last month? | Yes | Allocation data, which most already have |
Why did consumption increase by 40% in one month? | Rarely | A connection to which application, team, and workflow drove it—meaning Business Architecture |
What value does the consumption create? | No | A connection to the business process that the consumption supports |
What happens if we roll this out across the entire department? | No | An architectural view in which consequences can be modeled before committing |
How Clariox helps
We use Ardoq to build the Business Architecture that connects the consumption you can already see in your cost tool to your systems, processes, and owners, so your finance and IT leadership can actually answer the three questions above. We do not deliver FinOps; we deliver the foundation upon which FinOps work stands.
Where it is produced.
We connect the cost taxonomy from your existing tool to the same architecture and business model in Ardoq, so a cost pool is no longer just a number, but is linked to the concrete applications and teams driving it. At the same time, Ardoq’s AI Lens provides visibility across the entire AI landscape, including AI features embedded in existing SaaS tools and AI adopted outside any formal process, which is typically a blind spot for both cost tools and operational platforms.
How it is consumed.
With the Business Architecture overview in place, it becomes possible for your organization to introduce token budgets, model selection standards, and context management policies that can actually be enforced because they are tied to a named owner and a concrete system, not a general statement of intent.
What value it creates.
Ardoq's scenario modeling makes it possible to test the impact of a decision before committing—for example, what it actually costs to scale a pilot project with 20 users to an entire department, instead of discovering it three months into the rollout.
We always start in the domain where the pain is greatest right now, rather than tackling the entire landscape at once, in line with the crawl, walk, run approach that the FinOps Foundation itself recommends for AI cost management.
💡 Frankly speaking: an architectural view does not replace the need for high-quality cost and operational data, and it is not a one-time exercise. Its value comes from being kept up-to-date. Moreover, governance mechanisms do not implement themselves just because visibility exists. Architecture makes it possible to enforce them meaningfully. The decision must still be made by people within the organization. |
Ready to answer why your AI bill increased, and what to do about it?
📅 Book a non-binding conversation at clariox.dk/contact
Sources: McKinsey, "Burning through the AI budget" (July 2026) · FinOps Foundation, State of FinOps 2026 · FinOps Foundation, Framework 2026 (Domains, Capabilities, Scopes, Personas, finops.org) · FinOps Foundation, the "Executive Strategy Alignment" capability · Tokenomics Foundation (Linux Foundation, announced as intent at FinOps X 2026, no published standards yet) · Optimum Partners, analysis of 2.4B enterprise API calls (2026) · Goldman Sachs, projection of token consumption growth towards 2030 · KPMG, Global AI Pulse · Flexera, State of ITAM Report 2026 · Deloitte, State of AI in the Enterprise 2026 · SAP LeanIX, Agentic AI Survey 2026.
Troels Rendbæk Sørensen - CEO & Founder