Why AI is raising worker productivity but not making the economy more efficient

Technologists claim AI will help optimize workflows and supercharge the U.S. economy’s productivity—a measure of how efficiently resources such as labor are being converted to goods and services. While that growth has yet to show up in the data, AI might be responsible for the discrepancy in productivity statistics so far. 

In certain professions, employees who use AI are more likely to produce the same amount of work in less time, potentially saving an entire workday a week, according to a study by the London School of Economics last year. Economists call this an example of capital deepening, or when workers gain access to better tools and their individual productivity rises as a result—like when a construction worker trades in a shovel for a mechanical excavator.

There’s another example of this process that might be more analogous to the age of AI, put forward in a research brief published Tuesday by the Federal Reserve Bank of San Francisco. Just as with companies spending lavishly on AI integration today, economists analyzing the first days of the Internet in the early and mid-1990s might have been similarly puzzled. Employees suddenly had access to groundbreaking technology, but many firms remained stuck in the trenches of a “productivity paradox” that plagued the U.S. between the 1970s and 1990s as massive investments in IT failed to translate to improved efficiency.

That lull proved to be just a lag, of course, and if history were to repeat itself, the U.S. economy might be in the early days of a historic productivity surge without even realizing it.

“Determining whether a prolonged period of high growth has begun or not is difficult in real-time and is usually only obvious with the benefit of some hindsight,” the Fed researchers wrote. 

Fickle productivity

There are two primary metrics economists use to gauge productivity, and the two are pointing in complete opposite directions. One is labor productivity, which measures output per unit of labor. The other is total factor productivity (TFP), a broader metric that encompasses how efficiently the entire economy is able to convert inputs into output.

Labor productivity has seen solid gains in recent years, but TFP has struggled to post significant growth since a post-pandemic surge. The Fed researchers interpreted the divergence as employees working faster and more productively on an individual level, but the workforce as a whole hasn’t necessarily become more efficient.

This pattern mirrors what happened during the computer and internet boom of the 1990s. Starting around mid-1996, labor productivity began accelerating more rapidly than TFP, but the full productivity benefits of the Internet didn’t materialize in the overall data until several years later.

The Nobel laureate Robert Solow encapsulated the dissonance with a quip that has since been immortalized: “You can see the computer age everywhere but in the productivity statistics,” he wrote in 1987.

A similar dynamic is playing out today, with commentators including Apollo’s chief economist Torsten Slok applying Solow’s framework to the AI age. Business investment in AI is surging because companies are forecasting a productivity boom, meaning each worker has access to a wider choice of tools that have yet to be efficiently integrated across the economy. 

The growing pains of AI adoption have been laid bare by multiple rounds of evidence. A Harvard Business Review study of 200 employees at a U.S. technology company published earlier this year found that employees who use AI tools did save time on their tasks, but that time was often redirected into other work resulting in fewer breaks overall. The end result was more time on the job for most workers, and a higher risk of burnout. A separate Harvard study found extensive AI use at work could lead to excessive cognitive loads, resulting in more cases of “brain fry.”

Another study by the Atlanta Fed from March was even more specific. The branch surveyed around 750 corporate executives and generally found productivity is improving thanks to AI. But perceived productivity gains, as reported by executives, were larger than what researchers could actually measure from indicators such as company revenue, which the Fed put down to “delayed output realizations.” 

Workers might feel as if they are becoming more productive with AI, and in many cases that could be true. But the lack of measurable impact for the economy at large comes with stark similarities to the early days of the Internet, when the data had yet to herald the imminent productivity boom.

“If today mirrors what we experienced in the mid-1990s, we may be in the early stages of a productivity boom driven by AI that will only become clear in retrospect,” the San Francisco Fed researchers wrote.