Stanford and ADP data show AI-exposed entry-level jobs contracting 3.8% a year

Workers ages 22 to 25 in highly AI-exposed occupations are contracting at 3.8 percent a year, accelerating from 2.8 percent in April 2024, according to the Stanford Digital Economy Lab and ADP’s Canaries Dashboard. The figures were surfaced on June 27, 2026.

The finding sharpens the debate over AI and entry-level jobs because the damage is concentrated by age, not spread across the workforce. Aggregate AI-exposed occupations shrank just 0.2 percent year-over-year; the contraction lands almost entirely on the youngest workers. Those ages 31 to 34 contracted 1.7 percent, while workers 35 to 40 grew 2 percent.

The dashboard draws on anonymized ADP payroll records covering 4.6 million workers across more than 730 occupations, about one in six American workers, updated monthly, according to the Stanford Digital Economy Lab. Early-career software developers and customer-service workers show some of the steepest employment declines.

Erik Brynjolfsson, who directs the lab, attributes the pattern to AI absorbing discrete tasks such as summarizing and formatting before it absorbs whole jobs, hitting the automatable functions that fill entry-level roles. “Whatever it is, it’s not going away,” Brynjolfsson said. The lab built the tool with ADP Research and launched it on June 10 alongside two companion trackers.

The data measures contraction rates, not zero hiring, and correlation is not proof that AI is the cause. Daron Acemoglu, the MIT economist and Nobel laureate, remains a prominent skeptic of large AI-displacement claims, cautioning that other forces such as interest rates and post-pandemic hiring corrections can drive the same numbers.

Whether the 22-to-25 contraction widens or stabilizes in the monthly updates will indicate if this is an AI signal or a business-cycle artifact.