The FRED^® Blog

The takeaway

More workers are using generative AI tools, using them more often, and saving more time because of them, according to new survey data in FRED.

The survey

Economists Alexander Bick, Adam Blandin, and David Deming built the Generative Artificial Intelligence Adoption Tracker using data from the Real-Time Population Survey (RPS). The RPS is a nationally representative online survey of working-age adults aged 18 to 64. Each quarter, employed respondents report how fast and how intensely they’re adopting AI technology both at work and outside of work.

The data

Our FRED graph above shows three data series from the generative AI adoption tracker: “Use Last Week for Work” (solid blue line), “Work Hours Assisted” (dashed green line), and “Time Savings” (dashed orange line)—all measured as a percent of employed adults (left axis) or percent of work hours (right axis).

All three lines trend steadily upward. 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, indicating adoption is broadening. The share of work hours assisted by generative AI grew from 4.1% to 6.3% over roughly the same stretch. Perhaps most importantly, the share of work hours saved thanks to AI assistance grew from 1.6% to 2.2%.

The summary

The information is based on how respondents self-report their “time saved,” so the measurements are inherently approximate. But these three data series together do tell a coherent story of AI adoption: More workers are trying generative AI, and the ones already using it are leaning on it more heavily. Rising time savings suggest AI is not just a novelty; it’s starting to free up real work hours. If this trend holds, generative AI could show up in broader productivity statistics in the years ahead.

How this graph was created: Search FRED for “Generative Artificial Intelligence, Use Last Week for Work: Employed Adults.” Click “Edit Graph” and select the “Add Line” tab. Search for “Generative Artificial Intelligence, Work Hours Assisted: Employed Adults” and click “Add Data Series.” Repeat the last step to search for and add “Generative Artificial Intelligence, Time Savings: Employed Adults.” Select the “Format” tab to customize line 2 and line 3 by selecting “Y-Axis position: Right.”

Suggested by Diego Mendez-Carbajo.

National employment growth in the second quarter of 2026 accelerated slightly to 0.3% compared with one year ago. However, this aggregate figure conceals wide variation in job growth across U.S. states: 38 states experienced job growth, while 12 states plus the District of Columbia experienced job losses relative to one year ago. The median state experienced job growth of roughly 0.4%. Moreover, 25 states experienced growth above the median, highlighting growing resilience.

Our FRED map above shows the percent change from a year ago in employment in each state during the second quarter. Minnesota and Nevada had the strongest growth, with net job gains of 1.4% and 2%, respectively. The District of Columbia and Montana had the largest net job losses, at -4.8% and -1%, respectively.

At the metro level, job growth on average displayed weaker trends than the state-level data, with the median MSA experiencing a slight job gain of 0.2% in the second quarter of 2026 relative to one year ago. The Waterloo-Cedar Falls, Iowa, MSA had the largest net job loss at -3.2%. In contrast, the Sandusky, Ohio, MSA had the strongest job growth, at 5%. These numbers tend to vary greatly from quarter to quarter, with even greater sampling errors than the errors at the state and national levels. So, be careful not to read too much into these data.

How these maps were created: Search FRED for “total nonfarm employees in Missouri” (or any other state). Click “View Map” and then “Edit Map.” Change the units to “Percent Change from Year Ago” and the frequency to quarterly with aggregation method “End of Period.” Under “Format,” select “User Defined Method” for how to group the data: Switch the number of color groups to 3 and change the colors to red for states that shed jobs (or a value less than or equal to 0), light yellow for states with modest job growth (or less than 0.4), and dark green for states with strong growth (or a value large enough to incorporate the rest of the states). For the second map, repeat the process with an MSA—St. Louis, for example.

Suggested by Violeta Gutkowski and Rehann Silvanus.

The takeaway

Employment of software coders has decelerated sharply since 2022. Recent research suggests this job-specific shock isn’t caused by a slowing industry but likely from the emergence of AI tools.

What employment data show

Our FRED graph above shows annual employment data for 1991-2025 reported by the US Bureau of Labor Statistics: The solid blue line shows the year-over-year percent change in total employment. The dashed green line shows the same measure but specifically for computer systems design and related services.

Annual employment growth in the computer systems design industry has outpaced overall employment growth, except for two distinct periods:

• In 2002-2003, the industry reeled back from the dot-com buildup of the late 1990s and was exposed to the 2001 recession.
• After 2022, a different type of employment shock took place.

What recent research shows

Analysis by Leland D. Crane and Paul E. Soto at the Board of Governors of the Federal Reserve System finds that aggregate employment of software coders decelerated sharply after the broad public launch of generative AI tools in November 2022.

The research points out that more than 30% of national coder employment is concentrated in the computer systems design and related services industry, 40% of which consists of coders.

These researchers were able to establish a likely causal effect between the public launch of ChatGPT and employment by modeling a counterfactual employment scenario without any external shock and then comparing that with the actual data. This allowed them to tell the difference between an occupational shock particular to coders and a shock to the broader industry that would also impact related computer system designs jobs.

To learn more about this topic, check this earlier FRED Blog post on the sharp decrease in job postings in the broader tech industry.

How this graph was created: Search FRED for “All Employees, Computer Systems Design and Related Services”. Click the “Edit Graph” button and select the “Add Line” tab to search for “All Employees, Total Nonfarm.” Don’t forget to click “Add data series.” For both selected series, change the units to “Percent Change, Annual,” with the aggregation method as “End of Period.” Last, change the date range to start in 1991-01-01.

Suggested by Maria Goffinet, Elena Roussanova, and Diego Mendez-Carbajo.