From ‘tokenmaxxing’ to ‘tokenpocalypse’: Companies rapidly backtrack after encouraging workers to spend with abandon on AI

According to leaked internal audio reported by 404 Media, consulting giant Accenture has begun discouraging employees from using AI for routine tasks, including turning PDFs into presentation slides, as it looks to curb soaring token costs. A token is the smallest unit of data that a large-language model can process at a time, and pricing depends on token consumption.

“We’re seeing from some of the data internally at least that it’s actually not our engineers that are driving the token consumption,” Justice Kwak, Accenture’s agentic AI strategy lead, said during the meeting. “It’s a lot of the non-engineers that are doing some of those behaviors.”

Moneywise reached out to Accenture for comment, but did not receive a response before publication.

AI spending spree

Accenture isn’t alone in trying to get a handle on AI costs. Gartner forecasts worldwide AI spending will reach $2.59 trillion in 2026, a 47% increase from a year earlier, even as many organizations struggle to demonstrate clear business returns from those investments.

As AI investment surges, companies are increasingly asking a different question: Is all that spending actually paying off? The “Tokenpocalypse” reflects that shift, as businesses that once encouraged employees to use AI for nearly everything begin setting limits and focusing on where the technology delivers the most value.

“The useful way to see it is that spending got ahead of proof,” Pragati Awasthi, an assistant teaching professor at Drexel University’s College of Computing & Informatics, told Moneywise.

Awasthi said lower prices have encouraged businesses to use AI far more often, offsetting the savings from cheaper tokens.While the cost of an individual AI token has fallen by nearly 90% since 2023, overall spending has still doubled.

“When a resource gets cheaper, people tend to use far more of it,” she explained. “Cheaper power bills rarely shrink once everyone leaves more lights on.”

Research reflects that shift. Awasthi claimed that industry data shows AI deployment costs have gone from a concern for just 3% of organizations in 2023 to 58% this year. Some companies have even burned through an entire year’s AI budget in just four months before introducing strict limits on employee usage.

Why companies are rethinking AI spending

The shift marks a reversal from the tokenmaxxing trend that encouraged employees to use AI for as many tasks as possible.

While consumers can typically subscribe to AI models for $10 to $200 a month, businesses often pay differently. Many AI providers charge enterprise customers based on token usage, meaning every prompt employees submit adds to the bill.

Think of it like the difference between sending a text and writing a novel. Simple requests, such as summarizing a meeting, use relatively few tokens, while more complex jobs like coding or building new software features can require tens of thousands.

Some companies have already begun putting limits on that spending. Bloomberg reported that Uber introduced a $1,500 monthly spending cap per employee for each agentic coding tool, including Anthropic’s Claude Code and Cursor. Employees can monitor their usage through an internal dashboard, while managers can approve higher limits when needed.

Using AI where it actually pays off

According to Awasthi, AI stops delivering value when companies measure success by how much employees use it rather than what it actually accomplishes.