Stanford/ADP Data: Entry-Level AI-Exposed Jobs Shrinking 3.8% Annually

New economic research from Stanford's Digital Economy Lab, drawing on payroll data from ADP covering 4.6 million US workers, confirmed this week that young workers aged 22-25 in AI-exposed office roles are losing job opportunities at 3.8% per year — while workers in the same age group in non-AI-exposed roles see 2% annual growth.

The Canaries Dashboard — named after the 'canary in a coal mine' concept — was built by economists Erik Brynjolfsson (Stanford) and Nela Richardson (ADP) and extends through April 2026, nearly four years of post-ChatGPT labor market data.

Key points:

• The 3.8% annual contraction specifically measures workers ages 22-25 in the most AI-exposed occupations: administrative roles, junior data analysis, entry-level copywriting, basic research, and standard customer service

• Entry-level employment in physical trades and non-digitized workflows is growing at 2% annually

• AI absorbs tasks before it absorbs jobs — the tasks it reaches first are those that do not require years of experience: retrieving, summarizing, scheduling, formatting, and basic information assembly

• A separate finding shows novice workers using AI see a 34% productivity boost, while expert workers see a 19% performance decline

• PwC's 2026 AI Jobs Barometer found the average wage premium for workers with specific AI skills has reached 62% above non-AI peers, up from 57% the previous year

• AI skill job postings are growing nearly 8 times faster than the overall job market

Why It Matters: The 3.8% contraction is not a temporary hiring freeze — it is a structural change in how organizations allocate junior professional labor. Organizations are skipping junior hires and assigning AI agents instead. The labor market is splitting into two tracks: AI-skilled workers whose opportunities expand, and AI-exposed non-skilled workers whose traditional entry points contract.