Early signs suggest that artificial intelligence is beginning to influence the job market, hinting at how it might shape the next economic downturn. Historically, companies have leaned on automation during recessions to cut costs. This time, however, the rapid rise of generative AI could change the traditional pattern of which workers are hit hardest.
JPMorgan senior U.S. economist Murat Tasci warns that white-collar knowledge workers—once largely insulated from recession-related layoffs—may face greater risks ahead. These “non-routine cognitive” roles, including scientists, engineers, designers, and lawyers, have typically been less cyclical and have often led recoveries after downturns. But Tasci says the next recession could be different.
Since the late 1980s, routine jobs—both cognitive (like sales and clerical work) and manual (like construction, maintenance, and transportation)—have steadily declined due to automation, and their recovery after recessions has grown weaker over time. In contrast, non-routine cognitive jobs have historically rebounded quickly.
Now, for the first time on record, workers from non-routine cognitive occupations make up a larger share of the unemployed than those from non-routine manual fields such as healthcare, personal care, and food service. Tasci calls this an “ominous” shift, suggesting that AI may already be eroding opportunities for these roles, especially at the entry level.
He notes that AI poses little additional threat to routine or non-routine manual jobs, which often require in-person physical work. But because white-collar knowledge workers now account for nearly 45% of U.S. employment—up from 30% in the early 1980s—a slowdown in their recovery could drag down the broader economy. “The jobless recoveries we’ve seen in routine work could repeat, this time driven by weak growth in non-routine cognitive occupations,” Tasci warns.
Not everyone shares this pessimism. Tech investor and White House AI and crypto advisor David Sacks recently pushed back on “doomer” views about artificial general intelligence. Writing on X, he argued that there’s still a clear division of labor between humans and AI: people must provide models with context, detailed prompts, and oversight.
In his view, predictions of mass job losses are exaggerated. “You’re not going to lose your job to AI,” Sacks said, “but to someone who uses AI better than you.”