Box CEO Aaron Levie wouldn’t be surprised if his engineers set up a tokenmaxxing leaderboard.
“Honestly, they might exist,” he told Business Insider. “I just don’t know that it’s gone super viral across the entire company yet.”
Levie wouldn’t be against it, per se. He said the trend — in which engineers race against each other to spend as many AI tokens as possible — is a “fun, novel thing” that leads in the right direction. It will push AI agents to their limits, he said, and find the biggest productivity gains.
Token rankings are the new hot-button topic in Big Tech. One Meta employee made a “Claudeonomics” leaderboard with titles like “Token Legend” before it was shut down, The Information reported. OpenAI also has a token leaderboard, according to The New York Times. A token is a measurement of computing that determines how AI use is priced; large language models break words into numerical inputs, treating each token as roughly ¾ of a word.
At the cloud storage company Box, Levie does track token spending, including at the employee level, but “we don’t celebrate tokenmaxxing in the same way,” he said. “But we are focused entirely on increasing the rate of product velocity and increasing the scope of our product roadmap as our major goal.”
He has other ways to discern which of his engineers are AI power users. Box has a Slack channel where people share their best practices for AI coding, Levie said, which already generally correlates with who uses agents the most.
The CEO wants them to use agents, and use them a lot. (Box sells enterprise AI agents.) It won’t only be engineers employing agents soon, Levie said; the tech will hit fields like marketing, finance, and law.
“You need to really start to figure out where can you get leverage from this new form of abundance of intelligence,” he said. “Tokenmaxxing is the extreme way.”
Allocating those agents (and their associated tokens) is another question. 90% of of the economy cannot tokenmaxx like Meta or a VC-fueled startup, Levie said.
“That’s the new frontier of things that enterprises have to think about,” he said.
Levie had heard several different strategies for token allocation. He described one unnamed company that had “Shark Tank”-style pitches for computing budget. (Levie declined to disclose the name of this company, as it was a Box customer.) Employees must ask for funding, test whether it works, and then scale it up, he said.
Other examples Levie said he had heard gave their most productive fields more powerful models, like Opus 4.6 and GPT 5.4. “You get more efficient and cheaper models down the stack through the organization,” he said.
Companies are “heat-seeking missiles” for productivity, he added. They’ll find the areas that need the tokens most.
But, no, Levie doesn’t plan to create his own corporate leaderboard and give out prizes.
“There’ll be a lot of hilarious outcomes that will exist around that idea,” he said. “You’ll have people spending budgets on completely ridiculous things.”
