Method & Data
How the durability model is built
Transparent assumptions, evidence links, and update notes.
Data last reviewed: 2026-07-21
Model assumptions
- Embodied work increases durability when tasks require on-site judgment and physical execution.
- Legal accountability raises replacement friction because a licensed human remains responsible for outcomes.
- Supply scarcity supports wage durability when qualified labor pipelines are constrained.
- AI risk is directional, not deterministic; scores are planning signals, not certainty.
Primary evidence sources
- Anthropic Economic Index (2025-26): measured AI usage and task coverage by occupation
- IMF, "New Skills and AI Are Reshaping the Future of Work" (Jan 2026): ~40% of global jobs exposed to AI-driven change
- Stanford HAI AI Index Report 2026 (labor chapters)
- Oxford Martin: The Future of Employment (2013), kept for historical grounding
Macro context references
- Stanford Digital Economy Lab, "Canaries in the Coal Mine?" (Nov 2025): 16% relative employment decline for workers 22-25 in the most AI-exposed occupations, from ADP payroll microdata
- World Economic Forum, "AI and the Future of Entry-Level Work" (Jun 2026): 37% of young workers globally — 69% in North America — hold jobs with medium-to-high AI exposure; construction and agriculture named least affected
- Challenger, Gray & Christmas (Jul 2026): AI cited in 101,743 US job-cut announcements in H1 2026 — the #1 stated layoff reason for four consecutive months
- Anthropic, "Labor market impacts of AI" (Mar 2026): observed AI usage already covers 75% of computer programmers' tasks, while ~30% of US workers in physical, in-person roles show effectively zero coverage. (Disclosure: Anthropic is an AI vendor.)
Data policy
Salary benchmark year: 2025. Role-level salary citations appear directly on each job page. Training links prioritize licensing boards, apprenticeship directories, and public workforce resources.
Changelog
- 2026-02-27 — Editorial decision board updated with scenario presets, baseline deltas, and role recommendations.
- 2026-02-26 — Category visual system and citation links normalized across directory and job cards.
- 2026-02-25 — 40-role dataset completed; salary citation mapping and training-link coverage expanded.