Tokens — the base units for measuring and pricing AI usage — have rapidly emerged as one of the most critical metrics for corporations. Enterprise AI model pricing has shifted from static subscriptions to dynamic, usage-based models, and now escalating AI consumption has transformed what began as a productivity experiment into a potential source of margin pressure. Because token charges typically rise with the volume of text processed and generated, higher employee adoption can quickly turn into a variable cost that is harder to forecast than traditional software seats. CEOs find themselves caught in a delicate balancing act.

AI is not the type of software tool that can simply be switched on and off at will. Over the past year, large language models have become deeply embedded across business processes, and every indicator suggests that this reliance is here to stay. Indiscriminately restricting AI use — an approach many executive teams are currently considering — would not only slow the growth and efficiencies companies have already achieved, but would also leave them less prepared to capitalize on the next generation of AI capabilities as experimentation becomes discouraged.

The danger is that executives fall into a trap of whipsawing their AI spending to balance the next quarterly budget, and in doing so, miss out on the fundamental transformation that AI will deliver when governed systematically rather than reactively.

The Second Phase of AI Adoption

The business world has entered the second phase of AI adoption, where the mandate has shifted from rapidly demonstrating competency and progress to proving that companies can extract the greatest possible value from AI without threatening their bottom line. This sustainable adoption phase will soon expose a divide with long-lasting implications across the corporate landscape — between organizations that can manage and scale AI economically, and those that cannot. Token costs are merely the first visible symptom of a broader governance problem.

Two key assumptions led to this inflection point. The first was an appealing but ultimately flawed thesis about what AI would enable businesses to do: swap human labor costs for model costs and capture the efficiency delta as profit. That tradeoff has proven far less straightforward. Many organizations that embraced this idea, reducing headcount in anticipation of AI-driven productivity gains, discovered they had shed institutional knowledge and critical engineering talent needed to effectively integrate and refine AI systems over time.

That realization is now striking companies that carried out large-scale layoffs. In one prominent recent example, Ford said it must rehire hundreds of engineers to address quality control issues arising from newly implemented AI tools. Executives directly cited the lack of veteran expertise as negatively impacting product development and limiting efficiency gains from autonomous systems.

The second assumption involved a misalignment between investors and management teams. In Teneo's most recent annual CEO and investor survey, 53 percent of investors expected a return on investment from AI within six months, while only 16 percent of large-cap CEOs believed they could deliver on that timeline. That deadline has now arrived. The gap matters because AI programs often require changes to workflows, data practices, vendor management, and employee training before benefits can be measured reliably, making short payback windows difficult to reconcile with how enterprise technology is typically adopted.

Five Recommendations for Boards and CEOs

The window to build disciplined AI governance remains open, but it will not stay that way indefinitely. Between the two firms, the authors engage with thousands of CEOs and boards globally to address these challenges. The following represents their current guidance.

First, reframe the AI conversation in the boardroom. Stop asking how much is being spent on AI. Start asking where AI investment is creating durable competitive advantages — and where it is generating consumption without compounding value.

Second, treat AI spend as a capital allocation decision, not an IT budget line. Usage that drives new revenue, creates differentiated customer experiences, or builds proprietary capabilities constitutes a growth investment. Usage that merely automates low-value processes is an operating expense. That distinction also gives boards a clearer way to compare AI initiatives against other strategic uses of capital, rather than treating all model consumption as one undifferentiated technology cost.

Third, establish governance mechanisms that match workloads to the least expensive model capable of performing them reliably. Employees will naturally gravitate toward the latest and most powerful models, even when older generations can produce the desired output. One practical implementation could involve a software layer that intakes prompts and automatically routes them to the appropriate model.

Fourth, reshape incentives around AI use. Employees should not be rewarded for using AI the most — an instinct many companies displayed early on, eager to demonstrate advanced adoption to investors. Nor should they be penalized with blunt usage caps, which can inadvertently stifle innovation. Instead, organizations should reward efficient AI use: achieving better business outcomes with the appropriate level of AI consumption. As noted in a Wall Street Journal analysis, the goal is not maximum usage, but maximum value per token.

Fifth, distribute AI governance throughout the organization. AI governance cannot be delegated to a single role such as a Chief AI Officer. Managers across all functions need to be accountable for guiding sustainable AI adoption within their teams. Employees, in turn, need practical support both to understand how their AI use is evaluated and to develop the skills required to meaningfully and efficiently contribute with AI over the long term.

The Path Forward

The companies that prevail in this phase of AI adoption will not be the ones that use AI the most, nor those that spend the least. They will be the ones that govern it best, consistently converting AI consumption into lasting economic advantage. Achieving that balance is difficult, particularly with a technology evolving at this pace. Yet getting that balance wrong may soon become an existential matter.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com.