But despite fears around how ChatGPT and other AI tools used in the workplace, AI might actually be helpful for workers.

A recent study by Erik Brynjolfsson, Danielle Li, and Lindsey R. Raymond shows that generative AI can actually boost productivity for workers. And the lowest-skilled workers could especially benefit.

“Our overall findings demonstrate that generative AI working alongside humans can have a significant positive impact on the productivity and retention of individual workers,” the authors wrote.

While their study was about generative AI and its impact on workers at a company, the researchers noted that the purpose of the study isn’t to “shed light on the aggregate employment or wage effects of generative AI tools.”

The working paper looked at data from about 5,000 customer support agents, the majority of whom work in the Philippines, at a Fortune 500 company. About 3,500 of those agents did not receive the AI assistance studied, while about 1,500 agents did. The paper stated that the AI system used was “designed to augment, rather than replace, human agents.”

To figure out changes in productivity, the researchers measured the number of issues agents resolved per hour. The AI system used by some of the workers in the study consisted of a GPT-based large language model along with machine learning algorithms. It was also trained on agent conversations.

The AI system helped out the customer support agents by providing suggested responses to customer inquiries. It also could provide the agent with links to internal documentation at the company. The authors noted that workers didn’t have to use what the AI system suggested.

The overall productivity increase for workers given the AI tool was about 14% on average, and the lowest-skilled workers saw an increase of 35%. Additionally, the report noted results for five groups of tenure — ranging from zero months to over 12 months. Like the results by skill level, the researchers found those with less experience were seeing the biggest productivity gains based on measuring issues agents resolved per hour.


Brynjolfsson, a professor at the Stanford Institute for Human-Centered AI and the director of the Stanford Digital Economy Lab, told Insider that high-skilled workers barely saw a productivity increase. He said that “it was close to 0%” for top workers.

“What’s happening is that the system is actually capturing a lot of the expertise, the tacit knowledge of those high-skill workers and making it available to the less-experienced workers, and that’s bringing them up,” Brynjolfsson said. “But as you can imagine, it doesn’t help the high-skill workers that much because that’s the stuff that they already knew.”

While the AI system wasn’t too helpful for the more-skilled workers, it had a robust impact on those who are relatively new to the company. The report stated that “it treated agents with two months of tenure perform just as well as untreated agents with over six months of tenure.”

The study didn’t just highlight productivity gains. Another finding from the study showed that the likelihood someone would leave the company plunged. The working paper stated that “on average, the likelihood that a worker leaves in the current month goes down by 8.6 percentage points.”

“It was really interesting to see the attrition numbers go down,” Brynjolfsson said. “Sometimes when people install a new piece of software, the fear is that they’re just driving their workers harder and squeezing more productivity out in the short run, but the workers hate it. That’s not what we saw here.”

The effects on attrition were especially evident for newer workers. Brynjolfsson said that overall, they found the customer support agents were happier.

“I think it was kind of a virtuous cycle where the call center operators were able to give better answers. That led to measurably higher customer sentiment,” he said, adding that the agents “ended up sticking around longer and they ended up liking their jobs more than they did before.”

Despite fears around job security because of AI, machine learning and generative AI can be helpful for people entering the job market, people who already have a job, and the employers that they work for.

“I think this is a real game changer in terms of not just productivity, but a new paradigm for doing software and capturing information,” Brynjolfsson said. “The fact that now, instead of people explicitly coding what they know and putting it into the system, the system can learn from data, learn from the chats, learn from experience, and then capture that knowledge and share it with other workers – that’s a really different way of creating software.”

While this study focused on customer support agents of a software firm, workers in all different kinds of positions have shared with Insider how they have used ChatGPT, an AI chatbot, to be more productive.

For instance, one person in the education sector said that ChatGPT “can be intimidating just because of how powerful a piece of software it is and just how suddenly it arrived in our lives.” However, he also told Insider that “I believe it’s for the better of education in the long run.”

And Brynjolfsson also believes that it’s possible AI could be helpful for high-skilled workers, despite the study’s results.

“In this particular setting, it was more beneficial for the low-skill workers, but it’s easy for me to imagine ways that high-skill workers would also benefit — where they can use it as kind of an aid that helps them brainstorm and be creative.”