The fact that an AI can do your job doesn't mean your job is actually going to be automated.
April 26, 2026
Original Paper
What Every AI Job Study Gets Wrong: A Framework for Predicting How Much AI Will Actually Transform Each Occupation
SSRN · 6422901
The Takeaway
Technical exposure is a poor predictor of how much an occupation will actually change in the real world. Institutional barriers and social norms create a massive gap between what a machine can do and what it is allowed to do. Many experts warn of a sudden wave of unemployment as AI capabilities expand. This framework shows that adoption is much slower and more complex than a simple software update. Most workers will see their roles evolve through a filter of human requirements rather than being replaced by an algorithm.
From the abstract
<span> <p>Existing research on AI's labor market impact focuses predominantly on exposure—measuring which occupations' tasks AI can technically perform. While valuable, exposure metrics alone cannot predict which occupations will actually experience transformation, because they ignore the institutional, economic, regulatory, and social barriers that mediate technology adoption.</p> <p>Massenkoff and McCrory (2026) empirically demonstrate that actual AI adoption remains a fraction of theoretical