Working paper

A Global Task-Skill Atlas of Automation Exposure

Jasmin Baier, Tommaso Crosta, Prashant Garg (2026)

Show abstract

We study how automation exposure changes when the same standardized tasks are evaluated in country-specific settings rather than under a single context-free benchmark. We build a task-level automation atlas for the full O*NET task universe and apply a common multidimensional labeling framework across 124 countries covering about 99\% of world population and GDP. The country-conditioned task layer shows large cross-country differences not only in exposed share, but also in technology channel, labour margin, and material AI use. A benchmark ladder shows that part of this variation follows a broad development gradient, while substantial country-specific deviation remains within income groups. We then carry those task-level results into skill, occupation, industry, and goods-trade reporting layers. The atlas is a descriptive measurement system for technical feasibility and mechanism; it does not estimate realized adoption, displacement, wages, productivity, or causal labour-market effects.

Cite: Baier, J., Crosta, T. and Garg, P. (2026). A Global Task-Skill Atlas of Automation Exposure.

Figures

Selected figures.

Task-skill network

How the task-skill network changes which skills appear most exposed.

When connections between tasks and skills are preserved as a network, the skill rankings shift — some skills look more exposed than a simple count would suggest.

Country maps

How much automation assists vs. replaces varies across countries.

Countries differ not just in how many tasks are exposed, but in whether those exposed tasks tend toward automation supporting workers or replacing them.

Country benchmarking

Exposure patterns track income levels — but with country-specific variation.

Higher-income countries tend to show different exposure profiles, but substantial variation remains within income groups. Descriptive, not causal.

Technology and trade

What countries import reveals how automation technology reaches them.

This view separates goods that support human workers from those that perform tasks directly, then compares countries by which type dominates their imports.

Supplementary

Supplementary materials.

Supplementary

How trade openness relates to direct-execution automation.

Countries with more exposure to substitution-type automation also tend to import more goods in the direct-execution category.