A large-language-model analysis of 18,797 work tasks across 124 economies found automation-exposed task shares ranging from 3.3% in the least-exposed economy to 61.6% in the most, rising with income but varying widely within income groups

The Global Automation Atlas study, authored by Prashant Garg, Tommaso Crosta, and Jasmin Baier, used a large language model to classify 18,797 work tasks across 124 economies. The study found that the share of tasks exposed to automation varied from 3.3% in the least-exposed economy to 61.6% in the most-exposed economy. This share generally increased with income but showed considerable variation within income groups.