By Two93 Staff Writer
In the spring of 2026, the head of technology finance at a large enterprise received a report that looked familiar. Cloud spending was on target. Software licensing costs were within forecast. Infrastructure budgets were behaving exactly as planned. Then someone asked a simple question.
“How much are we spending on AI?”
The room went quiet. The company had purchased enterprise licenses for GitHub Copilot, Microsoft Copilot, Google Gemini, and several AI-powered development tools. Adoption was accelerating. Internal teams were celebrating productivity gains. Executives were showcasing AI success stories at town halls. Yet nobody could answer a surprisingly basic question.
How much AI was actually being consumed?
More importantly, what was driving the cost?
The problem wasn’t a lack of data.The problem was that organizations were looking for AI costs using frameworks designed for a different era.For nearly two decades, enterprise technology spending followed a familiar pattern. Organizations negotiated multi-year software agreements, forecasted seat growth, managed renewals, and tracked utilization through licenses and subscriptions.
Cloud computing disrupted that model by introducing consumption-based pricing, which ultimately gave rise to the discipline now known as FinOps. Enterprises learned how to monitor compute, storage, and network usage and connect those resources to business outcomes.
Artificial intelligence is creating a new challenge entirely.
Unlike cloud infrastructure, AI consumption is not primarily driven by machines. It is driven by human behavior.
A single employee can consume dramatically more AI resources than hundreds of casual users. A software engineer running autonomous coding agents may generate thousands of times more tokens than an employee using AI to summarize emails. A business analyst experimenting with reasoning models can unknowingly trigger costs that dwarf traditional SaaS usage patterns.
The result is a growing blind spot between technology, finance, procurement, and vendor management teams. Nobody owns AI consumption. And increasingly, nobody understands it.
The Rise of Token Economics
A new discipline is beginning to emerge around this challenge.
Some practitioners are calling it Token Economics. While the term has roots in blockchain and cryptocurrency markets, its meaning inside the enterprise is evolving. Today, token economics refers to understanding how AI systems consume resources, how those resources translate into cost, and how organizations can govern that consumption effectively.
The growing body of research and industry analysis from firms such as Token Economics highlights an important reality: Tokens are becoming the fundamental unit of value exchange in AI.Every prompt, every agent interaction, every code generation request, every document analysis, and every reasoning task ultimately translates into token consumption.
The challenge for enterprises is that token consumption behaves very differently than software licensing.Licenses are predictable. Tokens are not. Two employees with identical software access may generate dramatically different costs based entirely on how they use AI. This creates a new management challenge that many organizations have yet to recognize.
GitHub’s Pricing Shift Signals What Comes Next
The industry’s direction became clearer when GitHub Copilot’s usage-based billing announcement introduced AI Credits as the mechanism for measuring and allocating AI consumption. On the surface, it appeared to be a pricing update. In reality, it represented something much larger. One of the world’s most successful enterprise software platforms was effectively acknowledging that flat-rate AI pricing is difficult to sustain at scale. The economics of inference, reasoning models, and autonomous agents increasingly require a consumption-based model. Many industry observers believe other enterprise vendors will eventually follow similar patterns. Not because they want to. Because the economics demand it. The same forces that transformed infrastructure pricing are beginning to transform AI pricing.
The New Organizational Gap – Who owns the token budget ?
Most enterprises already have clear ownership structures. Technology teams own platforms. Finance owns budgets. Procurement negotiates contracts. Vendor management governs supplier relationships.Security manages risk.
But AI introduces a category that falls between all of them.Who owns the token budget? Who decides when employees should use premium reasoning models? Who governs autonomous agents capable of generating millions of tokens per month?Who determines whether an AI workload should use a high-cost model or a lower-cost alternative?
In many organizations, the answer is currently nobody. That governance vacuum is becoming one of the largest financial risks in enterprise AI adoption.
Why Traditional Vendor Management Isn’t Enough
Vendor management teams have historically focused on pricing protections, service levels, renewals, audits, and contractual compliance. Those disciplines remain critical. But AI introduces a new layer of complexity. The most important cost drivers are no longer written into the contract. They emerge from daily user behavior. A contract may specify a price per token. It cannot control how many tokens employees consume. A contract may define AI credits. It cannot determine whether engineers route workloads to an efficient model or an expensive reasoning engine.
As a result, vendor governance must expand beyond contract governance. Organizations need consumption governance. They need visibility into usage patterns. They need policies around model selection. They need forecasting frameworks capable of estimating future token demand. And increasingly, they need leaders who understand the economics of AI at the same depth previous generations understood software licensing and cloud infrastructure.
The Next Evolution of FinOps or something New ?
Cloud FinOps helped enterprises understand the economics of infrastructure. AI requires a similar discipline, but one focused on behavior, models, agents, and token consumption. Some organizations are beginning to build dedicated AI cost management programs. Others are extending existing FinOps teams.
Many are turning to vendor management organizations because they already sit at the intersection of technology, finance, procurement, and governance.Regardless of where ownership ultimately lands, one thing is becoming clear. The next generation of enterprise governance will not be measured solely by contracts signed or licenses negotiated. It will be measured by how effectively organizations understand, manage, and optimize the flow of tokens through their business. The companies that develop this capability early will gain a significant advantage. Those that do not may discover that their greatest AI challenge isn’t adoption. It’s economics.
