FRED tracks generative AI workplace time savings

The St. Louis Fed's FRED Blog put a number on how much time generative AI is actually saving at work, and it is small but rising: self-reported hours saved climbed from 1.6% to 2.2% of work hours. Read this as a research blog summarizing survey data, not a Board policy move, a rate decision or an FOMC statement. The figures are self-reported, so treat them as a direction of travel rather than a precise measurement.

What the data is, and who built it

The post summarizes the Generative Artificial Intelligence Adoption Tracker, built by economists Alexander Bick, Adam Blandin and David Deming from the Real-Time Population Survey, a nationally representative online survey of working-age adults aged 18 to 64. Three series move together and all trend up. The share of employed adults who used generative AI for work in the past week grew from 28.2% in Q3 2024 to 39.2% by Q2 2026, so adoption is broadening. The share of work hours assisted by generative AI rose from 4.1% to 6.3% over roughly the same stretch, so the people using it are leaning on it more. And the headline series, the share of work hours saved, went from 1.6% to 2.2%. FRED is blunt about the caveat: the numbers rest on "how respondents self-report their 'time saved,' so the measurements are inherently approximate."

Why a finance desk should care about 2.2%

The money question behind the AI capex cycle is whether all the spending on chips and data centers ever shows up as measured output, and this is one of the first credible reads on that conversion. A move from 1.6% to 2.2% of hours saved is a labor-productivity signal at the level of the worker, not a stock-market narrative, and it is still far too small to explain the run-up in AI infrastructure budgets that names like SK Hynix's $4B Indiana HBM packaging plant represent. FRED draws the optimistic line itself: "Rising time savings suggest AI is not just a novelty; it's starting to free up real work hours," and "if this trend holds, generative AI could show up in broader productivity statistics in the years ahead." The words that matter for anyone underwriting the trade are "could" and "in the years ahead."

The takeaway

Use the tracker as a leading indicator, not a verdict: 2.2% of hours saved is the gap between the AI productivity story and the AI productivity data, and it is the number to watch quarter over quarter to see whether capex is converting. If the hours-saved series keeps climbing while adoption plateaus near 39%, that is the healthier signal, because it means existing users are getting more out of the tools rather than the count simply growing. Bookmark the RPS tracker in FRED and check it against your own portfolio's AI spending assumptions each quarter; a self-reported 2.2% is encouraging, but it is not yet the aggregate-productivity print that would justify the current pace of buildout.

For related coverage, see our reporting on the New York Fed's stablecoins and Mundell-Fleming analysis, BitGo's buy of NYDIG's trading arm, and SK Hynix's $4B Indiana HBM packaging plant.