Enterprise leaders are discovering that measuring AI success by token consumption alone creates perverse incentives, encouraging employees to rack up massive bills without delivering real productivity gains. A phenomenon known as "tokenmaxxing" has emerged at major companies including Amazon, JPMorgan, Meta, and Disney, where gamified leaderboards reward high AI usage regardless of actual business outcomes.

What Exactly Is Tokenmaxxing and Why Should IT Leaders Care?

Tokenmaxxing occurs when companies track employee AI adoption primarily through token usage, the computational units consumed when interacting with large language models (LLMs). Tokens are essentially small chunks of text that AI systems process; a typical sentence might consume 15 to 20 tokens. When companies create leaderboards celebrating top token consumers, employees begin optimizing for volume rather than value.

The consequences have been dramatic. One Disney employee interacted with Claude AI 460,000 times over a nine-day span, according to Business Insider reporting cited in the source material. In some cases, top token users at companies have reportedly spent millions of dollars on AI infrastructure costs. These aren't isolated incidents; the pattern reflects a fundamental misalignment between how companies measure AI adoption and what actually drives business value.

"Token leaderboards come from good intentions, a genuine desire to track how employees are interacting with AI tools. They're just trying to understand how people are using these tools, how many people are using these tools," said Trevor Stuart, senior vice president at software development support vendor Harness.

Trevor Stuart, Senior Vice President at Harness

The problem is that token usage alone tells an incomplete story. An employee might consume massive amounts of tokens by using expensive frontier AI models for simple tasks that could be handled by cheaper, simpler tools. It's like using a premium power drill to hang a picture frame when a hammer would suffice.

How Can Companies Measure AI ROI Without Creating Perverse Incentives?

The solution requires a more sophisticated approach to metrics that ties AI adoption directly to business outcomes. Rather than celebrating raw token consumption, organizations need to establish measurement frameworks that connect AI usage to tangible results.

For legal departments, which have been among the earliest adopters of enterprise AI, the challenge mirrors what IT leaders face across industries. Legal teams are using AI tools like Microsoft Copilot, ChatGPT Enterprise, and Google Gemini for contract review, document summarization, and regulatory research. These tools are delivering real efficiency gains, but translating those gains into budget-level proof points remains difficult.

"The big issue is that token use doesn't necessarily lead to productivity. Companies are using the number of tokens consumed as a proxy for how productively employees are using AI. The employees are de facto incentivized for using tokens or, in some cases, punished for not using enough tokens, and obviously, it's a metric that's very easy to game," explained Logan Wolfe, partner in the global enterprise transformation, AI, and sovereign tech strategy practice at Kyndryl.

Logan Wolfe, Partner at Kyndryl

Wolfe compares token usage metrics to rewarding software developers based on lines of code written, which historically led to bloated applications full of unnecessary complexity. When token usage becomes the key performance indicator (KPI), organizations inadvertently incentivize output volume over outcomes like efficiency, quality, and risk reduction.

Steps to Build a Balanced AI Metrics Framework

• Establish baseline measurements: Before deploying AI tools, capture current time-per-task for the work you intend to automate, current spending on outside vendors for those same tasks, and volume of work in those areas over the prior three to six months. These numbers don't need to be perfect, but they must be defensible.
• Track output quality and production impact: For developers using AI assistants, measure not the number of lines of code written, but the number of lines that actually made it into production. Wasted code that gets rejected or never ships represents wasted dollars that should be tracked separately from productive token consumption.
• Connect efficiency gains to budget outcomes: Productivity improvements that don't appear in the budget are invisible to finance and leadership. The most effective organizations connect AI-enabled efficiency directly to deflected outside counsel spend, headcount avoidance, or matter cost reduction.
• Run structured pilots with defined endpoints: Pilot programs lacking clear measurement frameworks rarely produce the financial impact data needed to drive broader investment. Define use cases, timeline, team, and metrics upfront, typically within a three-month window.
• Monitor cost optimization alongside consumption: Track optimizable dollars, wasted dollars, and tokens consumed together. Understanding which token usage represents genuine productivity versus which represents unnecessary spending is essential for controlling AI infrastructure costs.

The stakes are particularly high given the economics of AI infrastructure. Reductions in price per token and price per inference appear nowhere on the horizon, in part due to rising energy costs. This means that token-maximizing incentives can actually create an inverse curve of unit economics and return on investment (ROI) for AI initiatives.

"If you walk two miles a day, but consume 5,000 calories, you're unlikely to improve your health," noted Itamar Friedman, CEO of AI code review provider Qodo, using an analogy to explain why token usage alone provides an incomplete picture of AI value.

Itamar Friedman, CEO at Qodo

In legal departments specifically, AI is demonstrating 40 to 60 percent efficiency gains in certain use cases like contract review and document summarization. However, these gains are often absorbed into growing workloads rather than showing up as visible budget savings. The structural gap between efficiency and budget impact means that without careful measurement, organizations may be getting real productivity benefits while appearing to get no return on their AI investments.

The path forward requires discipline and clarity about what success actually means. Before selecting an AI tool, organizations should define exactly what problem it solves, who owns the outcome, and how success will be measured. This prevents the tool fatigue that occurs when deployment precedes strategy, locking teams into platforms that become outdated or economically unfavorable.

As AI adoption accelerates across industries, the companies that will see genuine ROI are those that resist the temptation to gamify consumption and instead build measurement systems aligned with actual business value. Token leaderboards might feel like a way to drive adoption, but they're more likely to drive costs through the roof while leaving productivity gains hidden in the budget.

" }