Which Workers Are Embracing AI?

Younger, less experienced workers were most likely to use ChatGPT, the researchers find. Each year increase in a respondent’s age corresponded to a 1 percentage point drop in the likelihood that the person used AI—and each year of experience made use 0.7 percentage points less likely. This aligns with previous research demonstrating that workers with more to learn on the job also have more to gain from generative AI.

However, despite their age and relative inexperience in the workforce, the employees most likely to use ChatGPT were also higher earners even before the technology was introduced. This indicates that higher-achieving people, particularly those with more education and better academic performance, may be more likely to use this technology.

The findings on the gender gap drew on a comparison of men and women in the same professions and the same workplaces, including people performing a similar mix of tasks. Women who reported facing an adoption barrier to ChatGPT use were the most likely to say it was because they lacked proper training. Men, on the other hand, were more likely to say they avoided using ChatGPT because of employer restrictions or concern about data confidentiality. Yet simply offering introductory materials to all workers may not suffice, the researchers write: Despite women’s stated need for more training, men were more likely in the study to sign up to receive information that described how to use ChatGPT in their jobs.

On average, the respondents estimated it would cut their work time in half for over a third of their tasks. At the same time, however, they were twice as likely to say ChatGPT provided smaller time savings for higher-skilled workers. And even among workers who said they thought ChatGPT could halve their time to perform a job task, 50–60 percent said they didn’t intend to use it.

When Humlum and Vestergaard directly informed some workers about experts’ assessments on the time ChatGPT could save them, the workers’ positive beliefs about time savings expanded—but a follow-up survey found that this didn’t increase the likelihood that they would actually adopt the technology.

The disconnect between workers’ belief in the benefits of ChatGPT and their adoption of it isn’t driven by “existential fears” of being replaced by generative AI, the research indicates. Only 10 percent of workers cited that as a concern. Instead, the findings suggest that employer policies are creating barriers. About 40 percent of workers said they needed more training, and 35 percent said employers actively restricted usage.

The researchers suggest that employers or government programs provide specific guidelines and targeted training on the technology. Workers who are lower achievers overall (lower earners with less education, even if they have more years of experience on the job) may need more help getting up to speed on ChatGPT. Women may also need additional targeting in order to narrow the gender gap.

Companies that restrict the use of generative AI may consider loosening the policies and reorganizing workflows around the tools, the researchers suggest. Then workers who understand the potential productivity gains from ChatGPT may be more likely to use it, which could help spur economic growth.