MW Even OpenAi's top economist thinks the AI job apocalypse is overstated. Here's why.
By Barbara Kollmeyer
Usage of AI is still lagging, even for most at-risk jobs, says study's author
OpenAI economists argue that AI exposure is a weak predictor of immediate job market pressure, with AI usage still lagging.
Pushing back at gloomier forecasts of AI job disruption, OpenAI economists have released a study showing that just 18% of U.S. jobs are at a "relatively higher risk" of being eliminated in the near term.
Among those, reorganization or expansion is more likely than outright job losses, according to the study from Alex Martin Richmond, labor economist at the AI research and deployment company.
"AI exposure alone is too blunt a measure to predict which jobs will face ?near-term disruption," Ronnie Chatterji, chief economist at OpenAI, wrote in an introduction to the study.
The report comes months after a viral Citrini Research report shook markets. The report laid out a doom-loop scenario of AI improvements wiping out white-collar employment and driving economic deterioration.
The U.S. jobs market, meanwhile, has been grinding along with little hiring and little firing, but with signs of an increasingly tough backdrop for college graduates. Concerns about AI disruption have hit software stocks hard this year, while large layoffs linked to AI have been seen in companies such as Oracle $(ORCL)$, Snap (SNAP) and Block (XYZ) in recent months.
The OpenAI study scanned more than 900 occupations covering 99.7% of U.S. employment, identifying which jobs will see the most pressure first, which could face reorganization and which could see more demand. The criteria included looking at where AI has capability, where humans remain essential, the response from demand to lower costs and actual use of AI.
The diagram shows the basis of the 18% figure, as well as one showing 46% chunk of jobs will see less immediate change:
The economist found that ChatGPT was used three times more in jobs most at risk of automation, but cautioned that higher exposure doesn't mean realized usage or short-term job-loss risk, nor is moderate exposure "ironclad protection."
Many occupations remain people-led, even if AI can do more underlying cognitive work. The riskiest segments included administrative support, data entry, bookkeeping, and editing, said the study. Reorganization is seen for legal services, accountants, financial analysts and technical writers, while those in software engineering, graphic design and creative arts could grow with AI. Unsurprisingly, healthcare, skilled trades, education and hospitality were identified as those expected to see less immediate change.
The study also looked at how much demand changes when price changes - demand elasticity - as a key part of figuring out job disruption from AI. The economist argued that, even if AI makes the cost of providing goods or services cheaper, demand and the need for more employees might go along with it.
AI automation can also cut the costs of tasks so dramatically and make output so abundant that more automation and job losses would be likely. "We see this as being a potentially critical part of the long-run effects of AI on the labor market, but it is unclear how quickly that will unfold," said Richmond.
The economist offered more proof as to why exposure alone to AI is "a weak predictor of immediate labor market pressure." Comparing the use of ChatGPT to human hours expected for a given task, the chart below shows that across every job category, current usage lags what is possible. Jobs with the highest automation risk also show the least realized exposure:
The economist said sectors at the highest risk are "clearest candidates for early-warning monitoring," and should get most attention from policymakers, such as data entry keepers and customer service reps, with workplaces possibly dwindling down to just a few employees.
While AI is linked to some employment changes, it's still too difficult to make clear ties between the two, though that could change rapidly. The chart shows how jobs believed to be at heavy AI disruption risk still aren't seeing that:
The study suggested that governments get out in front of job losses by investing in "better occupational measurement" to help the public better understand and evaluate AI risks to employment. Jobs at higher risk should be candidates for "early warning systems" and transition assistance. Those in the reorganization camp will need policymaker support, with guardrails to protect workers, while those jobs that grow with AI need policies that help expand access to those jobs.
Read: A bullish indicator for software stocks just flashed. Why a true comeback could be in the cards.
-Barbara Kollmeyer
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
April 17, 2026 08:58 ET (12:58 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments