Most job taxonomies start with titles and descriptions. TaskNet starts with interviews. Stanford Digital Economy Lab and NBER have released TaskNet as NBER Working Paper 35987, with a project page dated 11/05/2026. It uses human and AI-assisted interviews to build occupational task inventories from workers at scale. In six occupations, it interviewed 132 workers and produced 202 task statements from 2726 reports, all accepted by O*NET Center analysts. Every statement was validated by 318 incumbents in 7923 ratings.
What is new is granularity. Against O*NET 31.0 for the sampled occupations, TaskNet covers 75% of the 118 published tasks, adds 79 tasks with no O*NET counterpart and refines 58 existing statements. That changes AI-exposure estimates and gives L&D a more current map of real work.
For instructional designers, this supports task-level objectives and assessment. L&D managers can prioritise emerging tasks instead of refreshing whole curricula. HR leaders should ask whether job architecture and skills frameworks are built from worker evidence or legacy titles.
The implication is not to copy TaskNet. It is to treat task inventories as living evidence, updated with the people doing the work.
Source: Stanford Digital Economy Lab, 11/05/2026.
Comments 0
Sign in to like, save or comment.
No comments yet. Start the discussion.