AI Work Index

Data Downloads

The complete dataset behind this index is open. Download, analyze, and build on it. MIT licensed.

Downloads

Versioned Snapshots

Historical scoring snapshots for tracking changes over time.

Version Date Occupations Download
V3.1 (Current) March 2026 562 JSON · CSV
V2 January 2026 562 Archived
V1 December 2025 562 Archived

Quarterly snapshots will be archived here as new scoring runs are published. Use snapshots to track band movers and score drift over time.

Methodology Version

Version: V3.1 (three-layer scoring: exposure, bottleneck, market resilience)

Data vintage: 2024 wages, 2024/2025 demand signals

Occupations: 562 SSOC-coded occupations

Sources: MOM Singapore, Felten et al. AIOE, Pizzinelli/IMF, Anthropic observed usage, SOL 2026, Jobs in Demand 2025

Full methodology →

Data Dictionary

Key fields in the dataset. See the methodology page for derivation details.

Field Type Description
ssoc string Singapore Standard Occupational Classification code (4-digit)
title string Occupation title from MOM classification
major_group string Major occupational group key (e.g., professionals, managers)
gross_wage_median number Median gross monthly wage in SGD (MOM 2024)
gross_wage_25th number 25th percentile gross monthly wage in SGD
gross_wage_75th number 75th percentile gross monthly wage in SGD
exposure number AI technical exposure score (0-1). Based on AIOE index mapping to SSOC via ISCO crosswalk
bottleneck number Human bottleneck strength (0-1). Higher = stronger human advantage from judgment, creativity, interpersonal skills
net_risk number Net displacement risk (0-1). Formula: exposure × (1 - bottleneck) × market_modifier
risk_band enum Categorical risk: very_low (<5%), low (<15%), moderate (<30%), high (<50%), very_high (≥50%)
augmentation number Augmentation potential (0-1). How much AI can enhance (not replace) this role
impact_type enum ai_leveraged | at_risk | stable | mixed — based on exposure and bottleneck thresholds
market.market_momentum number Employment growth trend (0-1)
market.occupation_scarcity number Labour shortage signal (0-1). Derived from SOL and Jobs in Demand lists
market.market_resilience number Combined market buffer (0-1). Higher = stronger demand protection
evidence.anthropic_calibrated boolean Whether Anthropic observed usage data is available for this occupation
evidence.anthropic_gap number|null Percentile gap: observed AI usage minus theoretical exposure. Positive = more adoption than theory predicts
evidence.sol_match string|false Shortage Occupation List match: "exact", "prefix", or false
evidence.jobs_in_demand_match string|false Jobs in Demand (MOM 2025) match: "exact", "prefix", or false
confidence.score number Overall estimate confidence (0-1). Combines crosswalk quality, market data, and source freshness
confidence.level enum high | medium | low — categorical confidence
stability.label enum stable | watch | sensitive — how much the risk band changes under ±5pt stress test

This data is released under the MIT License. Attribution appreciated but not required.

Questions? See methodology or about.

Structural AI exposure scores, not employment predictions. Methodology

V3.1 · 2024 wages · Q3 2025 full + Q4 2025 advance labour data · 562 occupations · 80 roles