Cespial/delegation-divide

An open index of AI-agent delegability for 433 ISCO-08 occupations and 146 countries (v2026.1).
Explore it at balamlabs.org/adi · Data: doi:10.5281/zenodo.23552918 · Published by Balam Labs
bibtex
@article{espinal2026delegation,
title = {The Delegation Divide: An Open Index of AI-Agent Delegability across Occupations and Countries},
author = {Espinal Maya, Cristian and Jim{\'e}nez Londo{\~n}o, Santiago},
journal = {arXiv preprint},
year = {2026},
url = {https://github.com/Cespial/delegation-divide},
note = {Preregistration: doi:10.17605/OSF.IO/9ZNTF}
@misc{espinal2026adidata,
title = {{ADI} v2026.1: Agent Delegability Index data, code and {MCP} server},
author = {Espinal Maya, Cristian and Jim{\'e}nez Londo{\~n}o, Santiago},
year = {2026},
publisher = {Zenodo},
version = {v2026.1},
doi = {10.5281/zenodo.23552918},
url = {https://doi.org/10.5281/zenodo.23552918}
## Headline
35.2 % of task time in world employment is at least partly delegable to AI agents (L2+) in 2026
(95 % CI 33.9–36.7); only 2.5 % can be handed over in full (L3,
95 % CI 1.9–3.2). 18,796 O\NET tasks, 433 ISCO-08 unit groups,
146 countries, 2.34 billion workers.
| Income group (World Bank FY2026) | Countries | L2+ (%) | L3 (%) |
| High income | 53 | 48.7 | 3.7 |
| Upper-middle income | 43 | 35.0 | 2.7 |
| Lower-middle income | 33 | 28.3 | 1.8 |
| Low income | 17 | 22.5 | 1.1 |
H1, the delegation divide (supported): delegable employment rises with income — 7.23 percentage points
per log point of GDP per capita (95 % CI 6.52–7.83, p
ADI rates every O\NET 30.2 task statement on a frozen, preregistered rubric of four levels — no meaningful role (L0),
augmentation (L1), partial delegation (L2), full delegation (L3) — plus four barriers (physical,
accountability, trust, access), with two language models (Claude Sonnet 5.5 primary, Claude Haiku 5.5 secondary).
Tasks are carried to ISCO-08 unit groups through a documented crosswalk and weighted by ILOSTAT employment into a
country atlas. Cross-listed econ.GN · cs.CY · cs.AI.
Paper: arXiv (pending) · source in articulo/
Preregistration: OSF, doi:10.17605/OSF.IO/9ZNTF (rubric SHA-256 in docs/RUBRICASHA256.txt)
Dataset: release/v2026.1/ (CSV + Parquet, data dictionary, checksums) · Zenodo doi:10.5281/zenodo.23552918 (v2026.1; all versions: doi:10.5281/zenodo.23552917) · Hugging Face Cespial/adi
Static JSON for agents: release/v2026.1/api/isco/ .json, release/v2026.1/api/country/ .json, release/v2026.1/api/index.json
Tools for agents: pip install adi-index · MCP server adi-mcp (lookupoccupation, countryprofile, topoccupations, cite)
Version: v2026.1
## For AI agents
Python: pip install adi-index → adi lookup 2631, adi country COL, adi cite (data bundled, every record carries citation).
MCP: pip install adi-mcp then claude mcp add adi -- adi-mcp (tools lookupoccupation, countryprofile, topoccupations, cite).