Using Generative AI, and Saving Time With It
A quarterly tracker of whether generative AI is actually saving anyone time. Nearly four in ten employed adults now say they used it for work in the past week, up from roughly three in ten in late 2024. The two lines on the right axis are the harder question: what share of work hours the tools touch, and what share they save. Both are rising slowly. It is an early series and time saved is self-reported, so treat it as something to watch rather than settled evidence either way.
What does it show?
Adoption is the easy number; hours saved is the one that would eventually show up in productivity statistics, and it is still small and early.
Methodology
Three series from the Generative Artificial Intelligence Adoption Tracker built by Alexander Bick, Adam Blandin and David Deming from the Real-Time Population Survey, a quarterly nationally representative online survey of working-age adults, distributed through FRED. The signal line is the share of employed adults who used generative AI for work in the past week. The two context lines, on the right axis, are the share of work hours assisted by these tools and the share reported saved. Read this as an early tracker rather than a settled measurement. The series begins in 2024 Q3, so there are only a handful of observations and no full cycle to judge against. Time saved is self-reported, which makes it approximate in both directions: respondents may not notice small savings, and may credit the tool for time they would have saved anyway. Hours assisted and hours saved are different quantities — using a tool for part of a task is not the same as finishing sooner. The survey covers workers rather than firms, so it complements but does not replicate the business-level adoption measured elsewhere on this dashboard.