Canaries Dashboard: AI-Exposed Jobs for Young Workers Drop 3.8%

When ADP's payroll data covering 4.6 million workers is sorted by AI exposure and age group, a pattern emerges that aggregate employment figures obscure entirely. As Fortune reports, workers ages 22 to 25 in highly AI-exposed roles have seen employment contract at 3.8% annually as of April 2026, while the same age group in the least-exposed occupations has grown at 2% a year. That divergence is the core signal tracked by the Canaries Dashboard, a tool built by Stanford economist Erik Brynjolfsson in partnership with ADP Research spanning 730-plus occupations and roughly one in six American workers.

The dashboard's name is deliberate: young workers in AI-exposed fields are the canaries in the coal mine, the first to register a disruption that aggregate employment numbers have not yet caught. The mechanism is relatively direct. AI absorbs the task-level work that entry-level roles are built around, retrieving, summarizing, scheduling, formatting, before it displaces jobs outright. Senior workers retain harder-to-codify, experience-based skills that offer more protection. ADP Chief Economist Nela Richardson put it plainly: "In the aggregate, AI's impact on jobs remains modest. But when AI's impact is measured by career stage, dramatic differences emerge."

Brynjolfsson stress-tested the findings against several alternative explanations. Interest rate sensitivity was ruled out because rate-sensitive occupations like construction show the lowest AI exposure in the dataset. The effect also persists when the entire tech industry is removed, which rules out the tech-sector overhiring story. The pattern holds across different remote work data slices as well. And the decline has deepened rather than stabilized: from 2.8% per year in April 2024 to more than 4% per year since, covering nearly four years of data through April 2026.

The honest caveat is that not everyone reads the same data the same way. MIT economist and Nobel laureate Daron Acemoglu represents a credible skeptical position, arguing AI productivity estimates remain overstated; Brynjolfsson acknowledges the two have been "going back and forth" looking for common ground. The dataset, while large, reflects ADP's payroll client base and may not fully represent sectors outside that sample.

The harder question the dashboard raises but does not yet answer is what happens to the current 22-to-25 cohort over the next decade. If entry-level AI-exposed work traditionally builds the experience-based skills that protect mid-career workers, and that work is disappearing, this cohort may not simply catch up as they age. Mid-career workers ages 31 to 34 are already contracting at 1.7% year-over-year, which is the next signal worth watching.