After its public launch on Nov. 30, 2022, ChatGPT’s promise inspired a series of wild predictions about how tools of its ilk may upend the jobs market. These ranged from prompting an end to the human-dominated era to transforming jobs as we know them in the coming years.

Attracting special attention was one prediction spearheaded by Christopher Pissarides, a Nobel Prize laureate and London School of Economics professor who specializes in labor economics and the impact of automation.

Back in April, Pissarides predicted that generative AI would enable employees to be more productive in their roles and therefore spend less time on them. In other words, with the help of a tool like ChatGPT a four-day workweek could become a widespread norm.

A few months on—as attention has turned toward safety regulations and other companies have launched rival generative AI platforms—Pissarides feels even more certain about his prediction.

“Now, I do believe that [ChatGPT] will improve the quality of work and probably improve productivity even more,” he said in an interview with Fortune, adding that his thoughts on a possible shorter workweek are “more justified.”

Pissarides is far from alone. A recent study by think tank Autonomy looked into how AI-driven productivity increases could usher in a four-day workweek in the U.S. and U.K.

Legendary investor Ray Dalio also alluded to the possibility of a shortened week as AI becomes more widely adopted, while JPMorgan Chase CEO Jamie Dimon says future workers will be at their desks just 3.5 days a week.

Still, while signs point to a growing likelihood of a shorter workweek the question of AI’s impact on pay remains unanswered.

Pissarides believes despite fears that fewer hours at work could result in a loss in pay, that won’t necessarily be the case.

“Pay will not fall because of improvements in productivity,” Pissarides said. “Things will be done faster with ChatGPT, for example, with AI. We’d be able to do in four days what we’re doing in five days in many professions.”

Cycle of fear

Pissarides, who won a Nobel Prize in 2010 along with two other academics for their research on the economic effects of unemployment, says earlier automation tech—like robotics—were also expected to yield better outcomes than relying solely on human labor.

“We’ve always believed that automation technologies would improve productivity—and they would improve sufficiently [to] be doing better with time,” he told Fortune. Still, automation tools have long awakened “alarmist” estimates of automation taking over a large chunk of jobs.

The same is happening today with AI, as forecasts point to job losses to the tune of 300 million globally, according to Goldman Sachs.

AI has awakened new worries as well. Fears about AI misuse have been raised by industry experts, including Geoffrey Hinton, who is known as the “Godfather of AI.” In May, he said it could be hard to prevent bad actors from using the tech for harmful purposes. Hinton, who previously worked at Google, also said AI could manipulate humans and potentially outsmart them.

But Pissarides thinks it’s hard to make such predictions about AI because a lot about the tech still remains unknown.

“AI is different because there’s a lot more uncertainty about where it’s going next [and] uncertainty about its capabilities,” Pissarides said. “It’s a lot more difficult to predict where AI is going because we don’t know in which direction inventors … are going to develop it.”

AI in the workplace: Make it a friend

While AI can be used for good and bad, many in the business world agree AI will likely be a useful co-pilot to human counterparts. Tech giant IBM’s CEO Arvind Krishna wrote in a Fortune op-ed earlier this year that AI could help “tackle the kind of tasks most people find repetitive, which frees up employees to take on higher-value work.”

And a year since ChatGPT was made available to consumers, Pissarides thinks that’s the reason it has gained as much traction as it has.

“It’s much easier to identify things that we do at work that are very close to the capabilities of ChatGPT,” he said.

In the past, it was also much more expensive and computing power-intensive to put AI tools to use, which made it harder for companies to adapt such technologies. But as ChatGPT helps bridge that gap in price and computation power, Pissarides says he’s seeing employers starting to be more optimistic about the tech as an assistant to its employees.

“It’s a happier work place … more satisfied people at work,” he said.

It could take a few years before this reflects in productivity figures—although Pissarides cautions that it will require other factors including greater tech investment and supportive policies before society reaps AI’s benefits in earnest.

It ultimately comes down to the type of tasks AI is entrusted with—a study by Boston Consulting Group found that performance was boosted by 40% for some groups assigned on a creative project, while plunging 23% on business problem-solving tasks.

Even though Nov. 30 marks an important milestone in the AI journey, the path forward is not set in stone, Pissarides points out.

“It’s still a matter of choice,” he pointed out, referring to how AI is used.