For the past five years, the output for non-farm businesses has increased by a sizable 2% per year, The Economist reported citing statistics from the Bureau of Labor Statistics. This is a marked increase from the 1% productivity growth per year that defined most of the 2010s, and a trend that has taken even Federal Reserve Chairman Jerome Powell by surprise. 

“I never thought I’d see this many years of really high productivity,” Powell said in a March press conference. 

Yet, while the hype around AI over the past several years makes it a logical candidate for the main driver behind the productivity boom, Nicholas Bloom, a Stanford economics professor who is known for explaining the Great Resignation of the early 2020s, says it’s more likely work-from-home policies since the pandemic are fueling the trend. 

In trials, researchers found working from home has led to increased productivity because workers are saving time on commuting and general office time wasting. Bloom wrote in a post on LinkedIn earlier this week that working from home supports business creation and increases the labor supply by allowing more people to participate in the workforce.

In an email to Fortune, Bloom also noted the data show “a clear post-2020 surge in productivity growth exactly when WFH [work from home] ramped up.”

Still, large companies are increasingly moving away from work-from-home and flexible work policies. Since the start of the year, Home Depot, Instagram, and automaker Stellantis have forced employees back to the office five days a week. They join Amazon, the new No. 1 company on the Fortune 500, which started its five-day in-office requirement last year. In total, employees at more than half of Fortune 100 companies are required to work fully in-office, according to a report by real estate company Jones Lang LaSalle (JLL). 

Companies argue in-person work increases collaboration, and Bloom agrees. Yet, he said, bringing people back into the office five days a week is counterintuitive, and has more to do with either preparing for layoffs or the personal preference of a CEO, which Bloom said skews toward typically older and male leaders.

A hybrid approach of two days in office and three days remote may actually be the best option for increasing productivity, he said. As Bloom puts it, workers spend most of their time in the office on administrative tasks like managing their email and participating in zoom calls. With three remote days, workers can do the bulk of this work at home. This leaves two days when most employees are in the office that can be fully focused on in-person collaboration, mentorship, and company culture.

“Coming into the office does absolutely help with collaboration, mentoring and culture, but you don’t need to be in every day to achieve that,” he said.

Yet, once firms pick a work schedule, they should stick to it, he said. Going from three days in person to five days in person, for example, “generates huge anger, churn and disruption.”

Though the use of AI has exploded since ChatGPT was released in 2022, the rise in productivity over the past five years likely cannot be fully attributed to AI. 

Only in the past year has AI arguably started to make an impact on the workplace. As of April, Gallup reported for the first time that half of all employed people said they use AI in their role a few times a year, up from 46% in the previous quarter. Only 13% of employees said they use it daily.

Still, Fed Chairman Powell added in his March press conference that AI will “certainly contribute” to higher productivity in the future. The evidence may be on his side. A study by the St. Louis Fed from last year found that the time which generative AI models save employees could translate to a 1.1% aggregate productivity gain.