Version History

JobsVsAI methodology changelog

A public historical record of analytical model updates, capability taxonomy revisions, and estimation framework releases.

Methodology Navigation
← Plain-Language OverviewTechnical Methodology →

Preliminary Estimate Methodology — 2026 Q3

August 2026
Active Production

Introduced a transparent, multi-tiered estimation methodology to provide provisional AI exposure and replacement risk indicators for 390 occupations undergoing full task analysis.

Key Methodological Changes

  • Deployed three-tier estimation framework: E1 (point estimates from high task coverage), E2 (bounded range intervals from partial task coverage), and E3 (cluster proxy mapping from closest analysed occupations).
  • Explicitly separated Preliminary estimates from Verified analyses across all public interfaces and sitemaps.
  • Established strict evidence-basis disclosures and confidence interval boundaries.

Rationale & Analytical Scope

Why: To provide users with directional career clarity across hundreds of searchable occupations while preserving strict methodological integrity and preventing provisional estimates from being mistaken for validated task analyses.

Scope: 390 preliminary occupation estimates published; 507 Verified scores remained completely unaltered.

Multi-Factor Occupational Scoring Model — 2026 Q3

July 2026
Active Production Baseline

Launched the production multi-factor scoring architecture, evaluating occupational tasks across 15 capability dimensions with geometric bottleneck modeling and four structural friction layers.

Key Methodological Changes

  • Expanded capability evaluation from linear proximity to 15 discrete structural dimensions spanning cognitive, perceptual, physical, and governance domains.
  • Implemented weighted geometric mean aggregation with critical bottleneck caps to prevent cognitive strengths from masking physical or legal impossibilities.
  • Integrated four structural friction layers: Environmental & Physical Dependency, Human Accountability & Trust, Adoption Economics, and Labour-Market Resilience.

Rationale & Analytical Scope

Why: Task automation cannot be modeled as a simple linear average; specialized physical requirements, legal liability, and economic friction represent real-world barriers that separate software capability from actual human displacement.

Scope: 507 verified occupations published under the versioned scoring pipeline.

← Return to Plain-Language Methodology