How We Calculate AI Replacement Risk
Overview
ReplacedByAI provides data-driven AI replacement risk scores for over 1,000 occupations. Our methodology combines three leading approaches to automation analysis to produce a comprehensive risk score from 0 (very low risk) to 100 (critical risk).
Data Sources
O*NET Database (v30.2)
The Occupational Information Network (O*NET) is the United States' primary source of occupational information. We use O*NET data for occupation descriptions, task statements, skill requirements, work activities, and abilities — covering 1,016 occupations across 23 industries.
Frey & Osborne Automation Probabilities
Carl Benedikt Frey and Michael A. Osborne's landmark 2013 paper “The Future of Employment” from the Oxford Martin School estimated the probability of computerisation for 702 occupations. These probabilities (0-1) serve as our baseline automation risk assessment.
Skills & Activities Analysis
We classify O*NET skills and work activities into “AI-vulnerable” categories (data processing, pattern recognition, routine tasks) and “AI-resistant” categories (social intelligence, creative thinking, physical dexterity, leadership). The ratio of these scores contributes to the final risk assessment.
Scoring Methodology
Our AI Replacement Risk Score (0-100) is calculated using a weighted combination of three components:
Frey & Osborne Base Probability
The foundational computerisation probability mapped to the occupation's SOC code, scaled to 0-100.
Skills Vulnerability Analysis
Compares the occupation's AI-vulnerable skills (mathematics, programming, data analysis) against AI-resistant skills (negotiation, social perceptiveness, leadership).
Work Activities Analysis
Evaluates whether the occupation's key work activities are automatable (data processing, monitoring) or require human capabilities (creative thinking, caregiving, physical work).
The combined raw score is then passed through a sigmoid function to produce a well-distributed 0-100 scale, ensuring meaningful differentiation between occupations at all risk levels.
Risk Levels
Very high automation probability. Most tasks can be performed by AI.
Significant automation risk. Many core tasks are AI-automatable.
Moderate risk. AI will change the role but not eliminate it.
Low automation risk. Most tasks require human capabilities.
Minimal AI risk. Strong human-essential skill requirements.
Limitations
- AI capabilities are evolving rapidly. Scores represent current technology trends and may change as new AI systems emerge.
- The Frey & Osborne dataset was published in 2013/2017 and may not fully account for recent advances in generative AI.
- Individual job security depends on many factors beyond automation risk, including employer-specific practices, local labor markets, and regulatory environments.
- Our task classification uses keyword-based heuristics and may not perfectly capture the nuance of every occupation.
Attribution & Licenses
O*NET data is provided under a Creative Commons Attribution 4.0 International License by the U.S. Department of Labor, Employment and Training Administration. Frey & Osborne automation probabilities are from “The Future of Employment: How Susceptible Are Jobs to Computerisation?” published by the Oxford Martin School.