External signal·Stanford SALT Lab·Jun 6, 2025·Yijia Shao, Humishka Zope, Yucheng Jiang, Jiaxin Pei, David Nguyen, Erik Brynjolfsson, Diyi Yang·15 min read
Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce
“which occupational tasks workers want AI agents to automate or augment”
Summary
A Stanford auditing framework that maps which occupational tasks workers actually want AI agents to automate versus augment, and how those preferences line up with current AI capability. Using audio-enhanced mini-interviews and a Human Agency Scale (H1-H5), the team built the WORKBank database from 1,500 domain workers and 52 AI experts across 844 tasks and 104 occupations. The headline tension: a disconnect between where AI investment and capability are heading and where workers actually want automation — a task-level "where agents land first" map rather than a blunt role-level replacement claim.
Predictions for the future of work
Reframes the displacement debate from "will AI replace this job" to a granular automation-vs-augmentation picture, predicting friction where AI capability outruns worker desire for automation. Positions task-level preference-vs-capability mismatch as the leading indicator of where agents are adopted first — and where they meet resistance.
Originally published by Stanford SALT Lab · Jun 6, 2025
Read the original at Stanford SALT Lab