A CEPR column based on representative German worker data finds that AI adoption is better predicted by comparative advantage, which weighs AI productivity against user costs and worker productivity relative to pay, than by technical exposure alone. An exposure-only model explains 25% of adoption variation, while allowing for user costs and worker heterogeneity raises this to about 60%. Projections suggest that reductions in user costs, not AI capability improvements, will drive most future adoption growth.
185 other tracked developments touching the same claims.