Right before a job gets fully automated, the human experts in that field actually see a weird, temporary spike in their pay.
March 26, 2026
Original Paper
Within-Task Competence Heterogeneity and the Shape of Technological Displacement: A Generalization of Task-Based Automation Models
SSRN · 6339898
The Takeaway
Standard economic models suggest automation slowly erodes human wages. This model reveals a 'Hump Theorem' where technology first makes human experts super-productive and highly paid, creating a false sense of security before the technology reaches a 'cliff' where the human role is suddenly and entirely displaced.
From the abstract
Acemoglu and Restrepo's task-based framework models automation as a threshold sweeping across a continuum of tasks, producing monotonic decline in labor's share. We show that this monotonic displacement result arises as the κ → 0 limiting case of a more general model in which automation thresholds sweep across distributions of human competence within tasks. We relax this assumption by modeling human competence on each task as a truncated normal distribution governed by a single parameter, κ = (s