MW AI isn't taking jobs, but it may be limiting pay growth, study finds

By Steve Goldstein

For all the fears that artificial intelligence will take jobs, the evidence so far is that even the most exposed industries have not turned to automation to replace workers.

That's the finding of Barclays analysts led by Mark Cus Babic who examined AI exposure across occupations, industries and types of workers, in the U.S. and in Europe, drawing on government data, private reports from a firm called Revelio as well as survey data from the Federal Reserve and the European Central Bank.

They find that less than 10% of tasks are both core responsibilities for workers and can be better performed using AI. And they find that occupations most exposed to AI are not always at risk of being substituted, giving the example of foreign language translators where the consequence of errors can be substantial, as well as the interpersonal component of their roles.

That leaves professions including proofreaders, mathematicians, typists and bill collectors at risk of replacement. They also identify which large U.S. employers are most exposed to AI - such as IBM $(IBM)$, Alphabet $(GOOGL)$ and Capital One $(COF)$ - and the service-oriented ones that are least exposed, including Starbucks $(SBUX)$, Darden Restaurants $(DRI)$ and Dine Brands $(DIN)$.

The Fed and the ECB studies find that about a quarter of companies, and a third of workers, report AI usage, but that usage is growing rapidly. Job openings data from Indeed.com show firms are increasingly posting vacancies that contain keywords related to AI but the share of overall postings is still small, at under 0.4%.

The Barclays team then assessed the effects of AI adoption on employment growth. Consistently, they found a positive relationship - that is, more AI exposure led to more employment growth, rather than job losses.

That's at odds with other studies, including one from Nobel Prize-winner Daron Acemoglu, David Autor and other economists, which found more highly exposed firms reduced their overall hiring. The Barclays team note the Acemoglu team studied an earlier period - between 2010 and 2018, versus between 2017 and 2024 - and that theirs was specific to generative AI.

The Barclays examination does find that AI exposure was linked to lower wage growth. At the occupation level, for every one unit increase in AI exposure, annualized wage growth slips as much as 0.74 percentage points.

The most recent reading of the Atlanta Fed's wage tracker shows median wages growing at 4.3% per year.

-Steve Goldstein

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

May 01, 2025 05:30 ET (09:30 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.