Methodology
How We Calculate
AI Job Risk
Our risk scores synthesize multiple peer-reviewed datasets and government labor statistics. No single model has the full picture — we combine them.
Data Sources
O*NET OnLine (ONET 28.0)
The US Department of Labor's Occupational Information Network. Provides detailed data on skills, abilities, work activities, work context, education requirements, and job zones for 900+ occupations using SOC codes.
Frey & Osborne (2013) — The Future of Employment
Oxford University researchers Carl Benedikt Frey and Michael Osborne estimated automation probability for 702 US occupations on a 0–1 scale. Their "how susceptible are jobs to computerisation?" paper remains the most-cited study in this space.
BLS Occupational Outlook Handbook
The Bureau of Labor Statistics publishes median annual wages and 10-year employment projections for hundreds of occupational groups. We use this for salary data and growth rate indicators.
AI Exposure Research (Felten, Raj, Seamans 2021)
This research develops an AI Occupational Exposure measure mapping AI application domains to occupational abilities, providing a complementary measure to automation probability.
Goldman Sachs Global Economics Report (2023)
Goldman Sachs economists estimated that generative AI could expose up to 300 million full-time jobs globally to automation or augmentation, with roughly two-thirds of current jobs facing some degree of AI exposure.
WEF Future of Jobs Report (2025)
The World Economic Forum's 2025 report projects that 85 million jobs may be displaced by AI and automation through 2030, while 97 million new roles emerge. Analytical and creative thinking remain the most valued human skills.
McKinsey Global Institute (2023) — The Economic Potential of Generative AI
McKinsey estimates generative AI could automate work activities that absorb 60–70% of employees' time today, accelerating the timeline of automation across knowledge work by a decade or more.
Scoring Methodology
Occupation Identification
Each occupation is identified by its Standard Occupational Classification (SOC) code — the US government standard used across all federal labor statistics. We use O*NET's detailed occupation data as our primary occupational taxonomy.
AI Exposure Score (0–10)
We synthesize multiple research sources into a single 0–10 AI exposure score. Higher scores indicate greater likelihood of significant AI impact. The score accounts for both software AI automation and physical robotics.
Scores are normalized from: Frey & Osborne probability (0–1), AI occupational exposure measures, and O*NET work activity data on computer interaction, information processing, and physical work patterns.
Replacement Type Classification
Each occupation is classified into one of four disruption vectors based on O*NET Work Activities and Work Context data:
- Software & AI:High scores on 'Working with Computers', 'Processing Information', 'Analyzing Data'
- Physical Robots:High scores on 'Performing Physical Activities', 'Handling Objects', 'Operating Vehicles'
- Hybrid:Significant scores on both computer work and physical activities
- Human-Centric:High scores on 'Assisting and Caring', 'Thinking Creatively', 'Social Perceptiveness'
Skills Vulnerability Analysis
Individual skills from O*NET are classified by their susceptibility to AI replication. Skills involving social perception, empathy, physical dexterity, and creative judgment are marked as AI-resistant. Skills involving data processing, computation, and structured information tasks are marked as AI-vulnerable.
Risk Timeline Projection
Risk projections over 5, 10, and 20 years are modeled using a simple compounding function based on historical AI capability growth rates (~15% annually for relevant capabilities). High-risk occupations compound faster than low-risk ones.
These are statistical projections, not predictions. Actual outcomes depend on policy, economics, adoption curves, and factors not captured in any model.
Why These Scores Might Be Wrong
Frey & Osborne methodology assumes full job automation
The original 2013 paper classified jobs as either automatable or not — a binary framing that has since been widely critiqued. Most economists now agree that tasks within jobs are automated, not entire occupations at once. This overstates displacement risk for many roles.
Adoption speed varies wildly
Even if a technology exists to automate a role, economic, regulatory, and organizational friction delays actual deployment. A radiologist in Germany operates in a different regulatory environment than one in the United States — yet our scores treat them identically.
New jobs are created alongside displacement
Historical automation waves (industrialization, computing) displaced jobs while also creating new categories. The WEF Future of Jobs 2025 report estimates 97M new roles emerging by 2030 — roles that don't exist yet and can't appear in our dataset.
Our dataset covers 91 occupations
The US Bureau of Labor Statistics tracks over 800 detailed occupational categories. Our 91-occupation sample covers the major categories but misses many niche roles. If your job isn't listed, the most similar available occupation may not be an accurate proxy.
Limitations & Disclaimer
This tool is for informational purposes only. AI job displacement is complex and context-dependent. An occupation's risk score does not determine your personal job security.
Factors not captured in our model include: specific employer technology adoption, geographic labor markets, regulatory environment, union protections, cost of automation relative to wages, and the creation of new job categories.
The Frey & Osborne study has been criticized for overestimating automation probability. Later studies (Arntz et al., 2016; OECD, 2019) find that only individual task components—not entire jobs—face automation, reducing displacement estimates significantly. The McKinsey 2023 report and Goldman Sachs 2023 analysis suggest a more nuanced picture: most roles will be transformed, not eliminated outright.
Do not make major career decisions based solely on this tool. Consult with career counselors, industry experts, and labor market specialists for personalized guidance.