Reid Hoffman backs tracking AI token usage for experimentation

Following Meta’s decision to shut down its internal “tokenmaxxing” dashboard, LinkedIn co-founder Reid Hoffman has publicly supported the practice of tracking employee AI token usage as a tool for organizational adoption. The concept of “tokenmaxxing”—defined as the practice of tracking employee AI token usage as a proxy for AI adoption—has recently gained traction in Silicon Valley. A token is a small chunk of data that an AI model processes when understanding a prompt and generating a response, and it serves as the unit used to measure AI usage and cost. Meta shut down its internal dashboard following news of an AI leaderboard leak. Many companies have begun tracking which employees use the most tokens to identify who is embracing AI tools, though engineers have debated whether this metric is a viable measure of workplace productivity.

Hoffman shared his perspective during an interview at Semafor’s World Economy summit this week, offering advice for companies adopting AI. He argued that tracking employee token spend is a good idea, even though token usage is not a perfect example of productivity. According to Hoffman, a LinkedIn co-founder and venture capitalist, “Here’s one of the things that is a good dashboard to be looking at — doesn’t mean it’s a perfect example of productivity, but… how much token usage are people actually doing as they’re doing it?” He emphasized that companies should encourage employees across all different functions to engage and experiment with AI. While some of these experiments will fail, Hoffman noted that this is acceptable because the goal is to have a wide variety of people using the technology collectively and simultaneously.

Beyond tracking metrics, Hoffman suggested that companies should foster AI adoption through regular, cross-functional check-ins to share learnings and identify successful use cases. He advised that organizations establish a weekly check-in for groups to discuss what new AI applications they tried for personal, group, and company productivity, as well as what they learned from those efforts. This approach, he noted, helps companies identify highly effective use cases that emerge from broad experimentation. He explained that these check-ins do not need to involve everyone all the time, but a group-level check-in can reveal amazing results as employees share their experiences.

Why it matters

Hoffman’s endorsement of “tokenmaxxing” highlights a shift in how leadership views AI adoption: prioritizing broad experimentation and usage volume over immediate, perfect productivity metrics.