Generative AI, Productivity and the Future of Work

October 08, 2025
SHARE THIS PAGE:

Since ChatGPT was released publicly in 2022, news reports have focused both on the promised benefits and on the potential risks to workers from generative AI. Generative AI, which encompasses large language models (LLMs), is a type of AI capable of creating text, images, video, audio, code and other media in response to questions and commands. (For a primer on AI, see this Page One Economics article.)

Some companies have dramatically restructured their workforce, implementing mass layoffs, in the expectation of dramatically higher productivity; so far, the results have been mixed. Yet it’s not only businesses looking to embrace the new technology. Nearly half of U.S. employees are using banned AI tools at work, according to one news report.

Still, how much time has actually been saved through the use of generative AI tools like ChatGPT, Office Copilot, Gemini and Claude? And what has been the possible impact on overall productivity? I spoke with Alexander Bick, a senior economic policy advisor at the St. Louis Fed, to gain insights into these questions.

Bick and two colleagues, Adam Blandin of Vanderbilt University and David Deming of Harvard University, have been tracking the spread of generative AI through an innovative nationwide online survey, the Real-Time Population Survey. Unlike other surveys, the RPS examines not just overall usage but also usage by occupations and by industry.

The economists were surprised to find in their initial study how quickly the technology was adopted relative to other earlier technologies. They found that in August 2024—less than two years after the introduction of this technology—nearly 40% of the U.S. population ages 18 to 64 used generative AI to some degree, both at work and at home. In comparison, the adoption rate was 20% for personal computers three years after their introduction.

Generative AI Adoption and Workforce Productivity

Bick and his colleagues found that these new AI tools appear to be helping workers perform their tasks more quickly, thus increasing their productivity. Among their findings:

  • On average, generative AI users reported saving time amounting to 5.4% of their work hours, or roughly 2.2 hours per week in a 40-hour work week.
  • The more people used generative AI, the more time they saved. A third of workers who use the technology every day reported time savings of at least four hours in a work week, while only about 1 in 10 workers who use it just one day a week reported saving that much time.
 

Of course, these findings focused on only the amount of time saved. Some workers might not have saved time but instead improved the quality of their work product by using these AI tools.

Besides tracking individual time savings, the survey also collected the occupation and industry of each worker, allowing the researchers to identify AI usage and time savings by key occupations and industries.

They found that AI usage—and the time savings it yielded—was highest among math and computer workers and within the information services industry. It was lowest among personal service workers and within the “leisure, accommodation and other” industry. (See figures below.)

 
 

Plugging these findings into an economic model, Bick and his co-authors estimated that generative AI use represented a potential 1.1% increase in U.S. productivity by the second half of 2024 relative to 2022, which was before ChatGPT and other generative AI tools became widely available. As a comparison, labor productivity in the nonfarm business sector rose 2.3% in 2024 and 1.6% in 2023.

Again, potential is the key word. Some employers may not realize how much time is being saved with this technology, or the workers may be using AI without their employer’s knowledge. In either case, workers may take advantage of saved time to ease up at work rather than jump to the next task. If enough workers do, the use of reported time savings to calculate productivity gains could lead to an overestimation.

The ability of workers to use the time savings to ease up will likely be a short-term effect, as generative AI becomes more common and measurable, Bick said.

“Sooner or later, firms will realize it, and they are just going to expect more output when people have access to these tools,” he said.

Meanwhile, the researchers’ analysis examines only the actions of workers, not those of firms. Firms may use the technology to fully automate certain tasks and lay off the employees who previously did the work. This automation will result in higher productivity, but since such gains can’t be estimated through worker surveys like the RSP, survey-based calculations may underestimate the productivity gains.

How Rising Productivity from Generative AI Will Affect Workers

Overall, when workers become more productive, their wages typically rise because they’re creating more value. They also end up working less because they’re able to produce more with fewer hours of labor.

“That’s something we’ve seen since the industrial revolution,” Alex Bick said. “We have become richer and are working less.”

For individual workers, the rub is how productivity gains affect the overall demand for labor in a firm or an industry. A firm that steadily increases productivity but sees little growth in demand for its products may end up reducing its headcount.

Sometimes even entire professions disappear because machines can do their jobs cheaper and more efficiently, Bick said, pointing to switchboard operators and typists.

“Of course, this is tremendously costly for these workers as they must to find a new career, potentially without being able to use many of their previous skills,” he said. “As a consequence, they may incur significant wage losses.”

Generative AI won’t affect all jobs equally. The type of jobs most affected will involve tasks that AI appears to excel in, such as those done by software engineers and translators, Bick added.

The Effects of Generative AI by Industry

Just like with occupations, individual industries will fare differently.

“Information systems and finance, for example, stand out as those industries with the highest adoption rates, while sectors that involve a lot of manual labor, such as construction or food and accommodation, show the lowest adoption rates,” he said.

Even in Bick’s own field, generative AI is having an impact. Creating mathematical models and running complicated statistical analysis are basic tasks performed by economists. AI has allowed such work to be done much quicker and cheaper.

“It’s making economists more productive, but it's not replacing us,” he said. “Maybe at some point it will get there.”

What gives Bick some comfort is to look at the impact of personal computers, which started their own revolution more than 40 years ago.

“We don’t know exactly how everything will evolve with AI, but I think the adoption of PCs is a good reference point,” he said.

While that technology displaced jobs in some industries, the workers in industries that quickly adopted PCs saw benefits.

“The industries that had high adoption rates early had very high adoption rates later. And they also saw the largest wage growth,” he said.

Imagining Future Careers with Generative AI

What will be key is preparing the workforce for upcoming changes, Bick noted. Some lessons from the past include the following:

  • Help people to learn how to effectively use the technology and take advantage of it.
  • Help people whose jobs may disappear to transition to new careers via training.
  • Focus future training and education on job tasks for which AI is not effective or applicable.

Economists prefer to peer into data rather than crystal balls, so Bick offered no predictions. Yet a world intensely shaped by AI does weigh on his mind.

“That’s something I’m really thinking about because my son is six,” he said.

Any child at that young age imagines a wide range of jobs for their grown-up self. One day his son wants to be a construction worker; the next day he wants to be a pilot, Bick said.

AI has made the calculus of careers much more uncertain. If AI can be successfully combined with robotics, allowing for the creation of robots that can learn tasks, even manual jobs can be replaced, he noted.

“So I’m really wondering what the world will look like,” Bick said.

ABOUT THE AUTHOR
Greg Cancelada

Greg Cancelada is a coordinator with the St. Louis Fed’s communications team and managing editor of the On the Economy blog.

Greg Cancelada

Greg Cancelada is a coordinator with the St. Louis Fed’s communications team and managing editor of the On the Economy blog.

This blog explains everyday economics and the Fed, while also spotlighting St. Louis Fed people and programs. Views expressed are not necessarily those of the St. Louis Fed or Federal Reserve System.


Email Us

Media questions