Welcome to the age of AI sprawl

Tokenmaxxing briefly became the buzzy AI phrase of the spring, but the trend is already sputtering. Amazon pulled its AI leaderboard after employees gamed the system, Palantir’s CEO likened tokenmaxxing to a porn addiction, and Duolingo reversed a plan to factor AI use into performance reviews.

Meta and AT&T have reportedly started curbing AI use as costs soar. The rush to adopt new agents and vibecode solutions has produced what many call AI sprawl: workers juggling a bewildering array of tools, burning through budgets on duplicate work, and spending time "botsitting" to make outputs usable.

A survey of 6,000 digital workers in the US, UK, and Australia by Glean's Work AI Institute found 77% of AI users engage with multiple programs weekly, a third use four or more tools, and 60% move the same prompts between tools when initial outputs fall short. Individually, workers say they save about 11 hours a week, yet only 13% report those savings have "significantly improved" company performance.

United States, United Kingdom, Australia

tokenmaxxing, ai sprawl, botsitting, ai leaderboard, palantir, duolingo, meta, at&t, glean survey, ai tools