AI Sprawl: Companies Struggle With Too Many Tools, Wasted Budgets

The tokenmaxxing era ends as AI sprawl begins

The brief craze for "tokenmaxxing"—racking up AI usage metrics to demonstrate innovation—has already crashed into reality. Amazon dismantled its AI leaderboard after employees gamed the system with pointless AI tasks. Duolingo reversed course on tying AI use to performance reviews. Meta and AT&T have begun restricting AI access as costs spiral.

What's replacing tokenmaxxing is arguably worse: AI sprawl. Workers now deploy multiple AI agents and tools in uncoordinated fashion, creating duplicate work, burning through expensive token budgets, and spending hours "botsitting"—feeding AI systems the context and edits needed to produce usable output.

New research from Glean's Work AI Institute surveyed 6,000 digital workers across the US, UK, and Australia. The findings reveal a fragmented landscape: 77% of AI users engage with multiple programs weekly, a third use four or more tools, and 60% shuffle the same prompts between different platforms when initial results disappoint. Workers report saving an average of 11 hours weekly through AI, yet only 13% say these gains have significantly improved company performance.

Why it matters

AI sprawl represents a critical inflection point for enterprise technology adoption. Unlike previous waves of workplace software that arrived through centralized IT decisions, AI tools have proliferated through individual experimentation—creating hidden costs, duplicated effort, and coordination breakdowns that threaten to undermine AI's productivity promise before organizations can capture its benefits.

The collaboration crisis

"The pressure to signal innovation by mere AI awareness, knowledge, appetite, is so strong, and it's leading us astray," says Kate Niederhoffer, head of BetterUp Labs. Few companies are answering "the big why" about AI adoption—clarifying what they're trying to accomplish and how these tools should serve organizational goals.

The consequences extend beyond wasted spending. Lee Senderov, chief transformation officer at travel platform Travelport, described one employee who consumed 160 times more tokens than the next-highest user over four days. When workers operate in silos, they duplicate colleagues' efforts without realizing it. "You've got hard costs, you're spending more money on tokens that you don't need to be spending," Senderov notes. "But you also have duplicative soft costs of just, we're wasting effort."

Previous BetterUp research found that AI-generated "workslop"—documents and presentations lacking proper oversight—erodes trust between coworkers. As individuals lean on chatbots for answers and generative AI for tasks that once required a colleague's expertise, the communal fabric of work frays.

From individual gains to enterprise strategy

Rebecca Hinds, head of Glean's Work AI Institute, frames the problem as a "tragedy of the commons." Workers optimize for individual productivity gains while inadvertently degrading team effectiveness and organizational resources.

The challenge now is translating individual AI benefits into coordinated workflows. Senderov says Travelport is experimenting with centralizing AI usage—identifying when multiple people tackle similar problems and encouraging collaboration while surfacing best practices across the enterprise. The larger the organization, the harder this coordination becomes.

Meta's recent trajectory illustrates the tension: the company laid off 8,000 workers last month while planning to increase AI spending by 60-87% this year. CEO Mark Zuckerberg has suggested individuals can now accomplish what once required entire teams, but this risks dismantling the institutional knowledge and coordination that make organizations function.

The details were first reported by Business Insider's Amanda Hoover, who covers the tech industry as a senior correspondent.