Research

The AI Adoption Gap: 9% Daily, One in Four Weekly

Based on Bick, Blandin, and Deming, NBER / Management Science, 2025

A research team from the Federal Reserve Bank of St. Louis, Vanderbilt, and Harvard has been asking US workers a direct question since 2023: do you use generative AI at work, and how often? Their 2025 answer is specific. 9 percent of workers use it every workday. About one in four use it at least once a week.

Both halves of that number matter. Weekly use by a quarter of the workforce means adoption is real and moving. Daily use at 9 percent means the deep-skill group is still small.

Faster than the PC, still early

The same researchers compared generative AI with the last two technologies that changed office work. It reached its current adoption level faster than the personal computer did, and faster than the internet did at the same point after launch. So the shift is real. It is also unfinished: most workers have tried the tools occasionally at most, and a small group has folded them into daily work.

What the gap looks like up close

Occasional use and working fluency are different skills. Asking a chatbot to draft an email is the first rung. Knowing which of your weekly tasks the tools handle well, which they fail at, and how to check their output is the part most people have not built yet. The survey data suggests that second group remains a small minority of the workforce.

What this means for you

If you use AI rarely or never, the measured picture is calmer than the headlines. You are in the majority, and the daily-use group is small enough that joining it is a realistic move, not a catch-up sprint.

The starting point is knowing which of your own tasks are worth handing to the tools. That depends on your actual week, not your job title.