ADI v2026.1 — The Delegation Divide: An Open Index of AI-Agent Delegability across Occupations and Countries (data, code and MCP server)

Agent Delegability Index (ADI) v2026.1. 35.2% of task time in world employment is at least partly delegable to AI agents (L2 or L3) in 2026; only 2.5% can be handed over in full (L3). Delegable employment rises with income, from 22.5% in low-income to 48.7% in high-income countries (the delegation divide: 7.23 percentage points per log point of GDP per capita).

18,796 O*NET 30.2 task statements rated on a frozen, preregistered four-level rubric (L0 no meaningful role, L1 augmentation, L2 partial delegation, L3 full delegation) plus four barriers (physical, accountability, trust, access) by Claude Sonnet 5.5 and Claude Haiku 5.5; aggregated to 433 ISCO-08 unit groups and weighted by ILOSTAT employment for 146 countries.

Contents: release/v2026.1/ (adi_isco, adi_pais, adi_tarea as CSV and Parquet; claims.jsonl with every number in the paper; raw model ratings; DATA_DICTIONARY.md; SHA256SUMS), static JSON API (api/isco/<code>.json, api/country/<ISO3>.json), Python package adi-index and MCP server adi-mcp.

Data, figures and rubric: CC BY 4.0. Code: MIT. O*NET task statements reproduced under CC BY 4.0 (USDOL/ETA). Preregistration: doi:10.17605/OSF.IO/9ZNTF.