Reid Hoffman thinks companies should be watching how many AI tokens their employees burn through — just don't mistake volume for value. Speaking at Semafor's World Economy summit this week, the LinkedIn co-founder and VC endorsed employee token tracking as a meaningful signal of AI adoption, even as Meta was busy quietly shutting down its own internal tokenmaxxing dashboard after its leaderboard leaked to the press.
What's new
Hoffman stopped short of using the Gen Z terminology, but the endorsement was clear: "How much token usage are people actually doing as they're doing it?" he told the summit, calling it a good dashboard metric while acknowledging it's not a perfect proxy for productivity. His key caveat — you need context alongside the numbers. High token usage from exploratory or failed experiments is fine, he argued, as long as companies pair the data with an understanding of what those tokens are actually being used for.
Why it matters
Tokenmaxxing — tracking which employees use the most AI tokens as a measure of AI engagement — has become a genuine flashpoint inside tech companies. Engineers have pushed back hard, arguing the metric is roughly equivalent to ranking people by how much money they spend, with no regard for outcomes. The debate got louder after Meta's internal leaderboard surfaced publicly, prompting the company to pull it. Hoffman's public backing gives the practice a high-profile defender, though his nuanced framing — experimentation counts, but context is required — is notably softer than the blunter leaderboard approach Meta was running.
What to watch
Whether other executives follow Hoffman's lead in publicly endorsing token tracking will be telling. The more interesting question is whether companies build the contextual layer Hoffman recommends — tying usage data to actual outputs — or just keep scoreboarding raw consumption. One is a management tool. The other is a way to make people look busy with AI.