The Irreplaceable Score Methodology

Executive Summary

The Irreplaceable Score is a 0–100 composite index that quantifies how well-positioned an individual is to thrive as AI reshapes knowledge work. It is computed from four weighted dimensions, each scored 0–100 against a transparent rubric anchored in primary research from Anthropic, the World Economic Forum, Goldman Sachs, McKinsey, Stanford HAI, and peer-reviewed labor economics (Eloundou et al., 2023).

The four dimensions are: (1) Personal AI Fluency — the respondent's actual use of AI; (2) Company AI Maturity — their employer's organizational readiness; (3) Industry AI Disruption — sector-level exposure based on observed task-level data; and (4) Role Amplification — how much the respondent's specific job can be leveraged (vs. displaced) by AI.

Each dimension is decomposed into 4–5 sub-criteria with explicit weights, five level labels (Spectator → Native, or equivalent), and example behaviors. Composite scoring produces a score, a ±confidence interval, and one of four readiness categories: Behind (<30), At Risk (30–50), Emerging (50–75), Leveraged (75+). The rubric is designed for repeatable LLM scoring, psychometric defensibility, and enterprise reporting.

The 4 Dimensions

Each dimension is scored 0–100 against a transparent rubric with sub-criteria, five level labels, and example behaviors. The four dimensions combine into a weighted composite described in Section 3.

Occupation / industry exposure reference (Dimension 3)

Derived from Anthropic's observed SOC-level usage (Claude conversations) and Eloundou et al. (2023) task-level exposure, cross-checked against Goldman Sachs industry cuts. Exposure is converted to a 0–100 Disruption Score (higher = more disruption / repricing risk).

The Composite Score

The four dimensions are not weighted equally. Weights reflect (a) the lever the individual controls, and (b) the prognostic power of each dimension in published research.

Formula

Industry Disruption (D3) is a risk score — a more-disrupted industry should lower the Irreplaceable Score unless personal fluency or role amplification compensate. Flipping it (100 − D3) makes the composite point in the correct direction.

Confidence interval

Each dimension is estimated from a small number of inputs with known variance. We treat the final score as a weighted sum of four independent estimates:

In practice this yields a ±5 to ±9 confidence band for most respondents. The band widens when dimension scores sit on level boundaries, consistency checks flag ambiguous inputs, or free-text responses are short / low-signal.

Peer benchmarking

Percentile is computed against a rolling cohort of prior respondents, segmented by (industry, role seniority, region). We report the percentile only when ≥50 peers are present in the triad; otherwise we fall back to industry-only, then global. Inspired by Item Response Theory scoring practice: rather than z-scoring raw totals, we percentile-rank within calibrated subgroups so comparisons stay fair across cohorts.

Readiness Categories

The Irreplaceable Score maps onto four readiness bands, each with a characteristic behavioral profile and a research-anchored 12-month outlook. Thresholds are soft — a score of 49 and 51 should be treated similarly; surfaces should show distance to the next band to encourage action.

Scoring Guardrails

The rubric is transparent, which means respondents can try to game it. We apply three classes of guardrails.

Gamability detection

• Fluency vs. output mismatch: D1 inputs claim “Native” use but free-text descriptions show generic AI vocabulary → cap D1 at Operator ceiling (75).
• Tool-list bloat: claiming 10+ tools used weekly without describing one concrete workflow → cap Tool Diversity (1b) at 60.
• Builder claims without evidence: “I build agents” with no artifact URL / repo / internal link → cap Builder Index (1d) at 60.
• Company-maturity inflation: Scaler/Leader claim but role inputs don't reflect AI integration → discount D2 by 15%.
• Keyword stuffing: repetition of buzzwords (“agentic, LLM, RAG, fine-tuned”) without concrete nouns → consistency flag + widen CI.

Consistency checks

• D1 internal: Tenure × frequency × tools should tell one story. A 3-month user claiming 10 tools and Builder maturity is inconsistent — widen CI and discount the outlier sub-criterion by 20%.
• D1 × D2: Builder/Native individual at an AI-Dark company → flagged as “misaligned environment” (often a future job-switcher signal; noted, not penalized).
• D3 × D4: A role Leveraged in an Existential sector is rare (solo AI-native operators). Requires concrete evidence in free text; otherwise D4 is capped at Amplified (90).

Floor & ceiling logic

• Floor of 20: every completed assessment scores ≥20. A raw 0 doesn't reflect reality — even AI-dark workers have some career optionality.
• Ceiling of 98 (v1): 99–100 is reserved for validated AI-native operators (shipped AI products, public AI work with traction). This keeps the top band credible and prevents self-scored perfection.
• Hard floor: D3 ≥ 75 and D4 ≤ 30 → Irreplaceable Score is capped at 45 regardless of D1. Industry-role fit dominates when the role is being actively eliminated.
• Hard floor lift: D1 ≥ 85 → minimum Irreplaceable Score of 50. A Native/Builder individual can always move, even from a bad sector/role.

Sources & Citations