The atlas offers country-specific automation measures for 18,797 tasks in 124 countries. This could be useful for governments and development organizations that currently rely on exposure scores built for rich-country labor markets. Evaluation should concentrate on the LLM-generated task labels, the country conditioning, the validation exercises, and what the measure can and cannot say about actual employment effects.
These are provisional prioritization judgments. Each explanation links the score to paper-specific evidence and, where relevant, the public-use search.
Country-specific exposure measures could improve decisions on reskilling, social protection, and labor-market monitoring where rich-country occupation scores transfer poorly.
No neglectedness score is stored for this paper. The public search nevertheless found no independent reproduction or institutional white paper using the atlas measures.
The rating reflects the paper’s early stage and the possibility of checking a new measurement system before it is widely reused.
The open data and code make the construction auditable, but the 2.33 million labels rely on LLM classification. Construct validity, prompt sensitivity, country conditioning, and validation against observed outcomes are central.
The project is visible and unusually reusable, but it remains an early arXiv paper without evidence of broad institutional or scholarly uptake. That supports a moderate score.
The atlas is easy to inspect and reuse, but evidence of influence is still limited to research indexing and light public discussion. No consequential institutional use was found.
Automation affects the labour content of work differently across different contexts. Yet, most existing exposure measures assign fixed scores to tasks or occupations, limiting comparisons of automation exposure across countries. We develop a task-based and country-specific approach to classify automation exposure across the world to disentangle labor-substituting from labor-augmenting automation, the relevant technology channel, and the material role of AI. Our measure spans 124 countries, generating an atlas of 2.33 million task-country labels for economies covering 99% of world population and GDP. We present five descriptive results. First, exposure is highly uneven, ranging from 3.3% of tasks in South Sudan to 61.6% in China, and rises strongly with income, although substantial variation remains within income groups. Second, across countries, exposed tasks are skewed towards substitution rather than augmentation, but low-income countries are disproportionately exposed to substitution, whereas middle-income countries are more heterogeneous. Third, less technologically advanced forms of automation account for more than half of exposed tasks in low-income countries but about one quarter in high-income countries; while other more complex channels generally rise with income levels. Fourth, AI tends to be less prevalent in simpler channels of automation, but also more prevalent in labour-substituting margins in lower income settings and to augment labour in higher income settings. Fifth, we find that females seem to be disproportionately more exposed to labour-substituting automation than males. Our methodology provides a basis for comparing automation exposure across development stages, linking it with cross-country data and allowing us to treat exposure levels, labour margins, technological channels and AI involvement as separate dimensions.
No structured evaluation-challenges field is stored.
This is a data-and-methods paper building a task-based, country-specific atlas of automation exposure across 124 countries, disentangling labour-substituting vs labour-augmenting automation and the role of AI — squarely in Unjournal's growing 'social and economic impacts of AI' area, and with a strong development/LMIC angle (low-income countries disproportionately exposed to substitution; gendered exposure). It's an early arXiv preprint with no peer review, so independent evaluation of the classification methodology and the construction of 2.33 million task-country labels would add real value. Main hesitations: prominence is moderate (arXiv, not-yet-widely-known authors), and the contribution is largely descriptive/measurement rather than causal or directly decision-informing, so it sits in the 'monitor' band rather than clear prioritize-now.
Cross-country automation-exposure measurement that is comparable across development stages is directly useful to organizations thinking about the labour-market and inequality consequences of AI in LMICs — ILO, World Bank, IMF, and development funders weighing reskilling and social-protection policy. The finding that low-income countries and women are more exposed to labour-substituting automation is policy-salient, though the paper stops at description rather than telling a specific funder what to do.
First-version arXiv preprint (v1), no peer review yet, and it introduces a novel measurement methodology — exactly the stage where independent evaluation of construct validity and classification choices is most actionable before the atlas is widely cited or reused.
Targeted public-paper search guided by the Coefficient Giving application discussion, manual source and thematic-fit verification, deduplication, and Codex subscription scoring
This pass follows the application discussion's risk-to-response framing: exposure estimates are inputs, not outcomes, and should be assessed alongside actual task content, adoption, infrastructure, institutions, service-trade exposure, and feasible policy responses. It deliberately includes competing estimates and early evidence on BPO and export-linked work, youth and expertise pathways, firm adoption, and frontline health care. LMICs are not treated as one labor market, and inclusion is not endorsement or a completed Unjournal team decision.
Coefficient Giving request for proposalsAnthropic Economic Index report · 72% match
Country-level automation exposure data directly tests whether AI gains concentrate in high-adoption rich regions.
The Economics of Transformative AI (Korinek lecture) · 58% match
Task-based exposure atlas informs where automation thresholds (labor complement vs displacement) sit by country.
Economic growth under transformative AI (Trammell & Korinek) · 45% match
Disentangling labor-substituting vs labor-augmenting automation bears on whether labor share falls toward zero.
The authors have made the atlas, methods, data, and replication code easy to inspect. We found some research indexing and light public discussion, but no major institution or policy report using the measures yet.
The project website makes the 2.33 million task-country labels explorable and provides paper, methods, and data views. This materially lowers the cost for policymakers and researchers to inspect and reuse the measure, but availability is not the same as demonstrated use.
The public GitHub package contains analysis data, source extracts, prompts, code, and checks of the numbers reported in the paper. This should make independent reproduction easier. We have not found one yet.
RePEc’s NEP-TID report included the paper in its May 2026 new-research roundup. This helps with discovery within the field. It does not indicate policy use or independent validation.
Targeted search checked 2026-09-01. Search scope: Targeted exact-title searches across the open web, institutional and author pages, news and public social-media results, EA Forum/LessWrong, and policy/white-paper contexts. Evidence records distinguish commissioning or report use from independent discussion, media attention, indexing, and post-publication policy use. A search miss is reported as uncertainty, not proof of absence.
The privacy-safe aggregate contains 1 current rating: 0 team and 1 public. The human mean is 90.0/100. The team has not made a final prioritization decision.
No written discussion is public. Private and team-only text is never copied to this page.