Stage 1
Tasks are scored under a common framework.
The atlas classifies standardized work tasks by exposure level, pathway orientation, constraints, and implementation requirements before any country aggregation happens.
Methods
How the atlas is built.
Country-conditioned task exposure: standardized tasks scored under a common framework, then placed inside country-specific settings before aggregation.
Stage 2
Those task scores are placed into country context.
Country-conditioned weighting determines how those tasks combine inside each country’s occupational and production structure rather than assuming one universal context.
Stage 3
The public atlas surfaces are built from that layer.
World View, Country Profiles, Task Finder, the skill layer, the case studies, and the downloadable bundles are all reporting views built from the same country-conditioned task object.
Measurement infrastructure
From source taxonomies to country-task measures.
Summary map. Switch to Detailed stages for the full pipeline.
Measurement framework
Four dimensions per task
Each task is evaluated on exposure level, dominant technology channel, whether automation tends to support or replace the worker, and whether current AI tools are materially integrated into the task today.
Benchmarks
Country-aware and simpler comparison frames
The main benchmark accounts for each country's specific task mix. A simpler context-free benchmark is available for comparison — it uses the same evaluation framework but ignores country-level differences.
Task-skill structure
Tasks and skills as a network
One task often requires multiple skills, and one skill often spans many tasks. The atlas preserves this network structure rather than flattening tasks to a single skill.
What the atlas measures
- Which tasks show high exposure to automation — software, robotics, or AI
- How exposure varies across 124 countries
- The network of connections between tasks and skills
- Which goods traded between countries are linked to automation technology
- Patterns in how automation relates to income level, region, and trade
What the atlas does not measure
- Whether automation has actually happened
- Job losses or displacement
- Effects on wages or productivity
- Organizational or policy effects
- Causal relationships of any kind
Limitations
The atlas measures task exposure, not outcomes.
Country-level estimates reflect task structure as measured under the atlas framework, not direct observation of local institutions or labor markets. The trade layer is a descriptive proxy based on goods classifications, not a measure of realized adoption. Read the paper for the full limitations discussion.