Reid Hoffman is wading into the debate over whether tracking AI token consumption actually tells us anything useful. His take? It's a decent adoption signal, but a terrible productivity metric on its own.
The discussion comes as more companies try to quantify their AI investments. Token counts are easy to measure, which makes them tempting as a KPI. But Hoffman points out the obvious problem: high token use could mean people are getting tons of value, or it could mean they're spinning their wheels with poorly designed prompts.
This matters if you're trying to justify AI spending to leadership or figure out if your team is actually benefiting from these tools. Raw usage numbers don't tell you if someone wrote better code, closed more deals, or just had a really long conversation that went nowhere.
Hoffman's advice is to pair token metrics with context. Look at what people are using AI for, what outcomes they're achieving, and how their work has changed. The number itself is just a starting point.
For anyone managing AI adoption at their company, this is a reminder that the easy metric isn't always the right one. Token tracking has its place, but you need the full picture to know if your AI investment is actually paying off.
