For top-tier pros, AI won't just slowly take your job—your value will explode for a minute and then fall off a cliff.
The paper finds that for cognitively complex jobs like law or medicine, the demand for human experts will rise superexponentially as AI tools make them more productive, only to suffer a discontinuous collapse the moment the AI reaches a specific competence threshold. It suggests 'Generation AI' faces an all-or-nothing employment future.
<div> Estimating Within-task Competence Heterogeneity (κ): </div> <div> A Cross-occupation Approach with Implications for AI Displacement </div>
SSRN · 6347218
The parameter κ = (s max-µ)/σ governs the shape of AI-driven labor displacement in Deobhakta (2026). When κ is small, displacement follows the monotonic stepfunction of Acemoglu and Restrepo (2018). When κ is large, displacement follows a spike-cli trajectory: per-expert value rises superexponentially before discontinuous collapse. This paper provides the rst cross-occupation empirical estimates of κ, using two complementary approaches: (1) within-occupation wage percentile ratios from BLS Occup