How We Calculate AI Risk Scores
Full transparency on our scoring methodology, data sources, and update process. Every score is auditable.
535
Jobs Analyzed
2172
Tasks Scored
30
Industries
535
Jobs with Citations
The Scoring Formula
Overall Job Risk Score
risk = weighted_average(task₁, task₂, ... taskₙ)
Where each task score (0–100%) represents AI automation potential
Each job is decomposed into 4–7 core tasks that represent the actual work a person does daily. Each task is independently scored on a 0–100% scale.
The overall risk score is the weighted average of all task scores. Tasks that consume more of the job's time receive proportionally higher weight.
Example: Web Developer (58% risk)
Calibration check: For 12 high-impact jobs (Professor, Pharmacist, Lawyer, etc.), we conducted deep manual research and expanded to 5–7 tasks with individually verified scores from published studies.
How Each Task is Scored
Five dimensions determine each task's automation score:
1. Current AI Capability
Can today's AI (LLMs, agents, computer vision, robotics) perform this task? Based on published benchmarks.
2. Near-Future Trajectory
Based on research trends, how likely is full automation within 5–10 years?
3. Task Complexity
Does it require physical dexterity, emotional intelligence, or creative judgment that AI cannot yet replicate?
4. Regulatory Barriers
Do licensing, safety, or legal requirements (FDA, bar exam, medical boards) slow AI adoption?
5. Real-World Adoption Rate
Is AI actually being deployed here, or only theoretically capable? Anthropic's 2026 study found actual AI usage is just a fraction of theoretical capability — we weight this heavily.
Research Sources
Anthropic — AI Labor Market Impact Study
March 2026Primary source. AI theoretical coverage exceeds 80% in several occupation groups. Computer/math and business/finance at 94.3% exposure. 16% employment decline for workers ages 22–25 in AI-exposed roles.
View study →Goldman Sachs Research
2023–20266–7% of US workers (~11M jobs) projected for long-term AI displacement. AI-related job losses at ~20,000/month in 2026. 300M jobs globally exposed. 25% of US work hours automatable.
View research →McKinsey — Agents, Robots, and Us
Nov 2025Current AI could automate 57% of US work hours. 40% of total jobs are high-automation roles. AI fluency demand grown 7x since 2023. 32% of companies expect AI-driven workforce reduction.
View report →Oxford Martin School (Frey & Osborne)
2013–2026The foundational framework for estimating automation probability of 702 occupations. Continuously updated with AI-specific capability assessments.
World Economic Forum — Future of Jobs Report
2025170M new roles created vs 92M displaced by 2030 (net +78M). 41% of organizations expect AI-driven reduction. 40% of skills will change by 2030.
View report →Federal Reserve Bank of Dallas
Feb 2026AI-exposed sector wages up 16.7% vs 7.5% national avg since 2022. Computer systems design employment down 5%. Total US employment up 2.5% since ChatGPT launch.
View research →Bureau of Labor Statistics
Feb 2026Official US employment data. 92,000 jobs shed February 2026. Unemployment at 4.4%. OEWS wage data used for salary benchmarks.
View data →Data Quality & Integrity
Every score undergoes multi-step validation:
Task Decomposition
Each job is broken into 4–7 constituent tasks reflecting daily work activities, verified against BLS O*NET task descriptions.
Individual Task Scoring
Each task scored 0–100% across 5 dimensions (capability, trajectory, complexity, regulation, adoption).
Consistency Validation
Automated checks ensure no job's overall risk diverges more than 10 points from its task average. 535 jobs pass this check.
Manual Expert Review
12 high-impact jobs received deep manual research with expanded 5–7 task breakdowns and individually cited scores (Professor, Pharmacist, Lawyer, Web Developer, and more).
Source Attribution
535 of 535 jobs have linked research citations. Each job references 4 peer-reviewed or institutional sources.
Risk Tiers
Very Safe (0–19%)
84 jobsRoles requiring deep human judgment, physical presence, or emotional connection. Minimal AI threat in the foreseeable future.
Low Risk (20–39%)
Some tasks automatable, but core responsibilities require human skills. AI augments, not replaces.
Moderate Risk (40–59%)
Significant portions automatable. Workers should actively upskill and diversify their capabilities.
High Risk (60–79%)
Most routine tasks automatable. Role will transform significantly. Career pivoting recommended.
Extreme Risk (80–100%)
79 jobsNearly all tasks automatable with current or near-future AI. Highest probability of significant reduction or elimination.
Update Process
Quarterly
Full score review cycle
Continuous
Breakthrough AI updates
Per-Job
Last reviewed dates tracked
Every job page now displays its last reviewed date and research citations. When major AI capability announcements occur (new model releases, regulatory changes, major studies), affected jobs are re-evaluated immediately.
Important Caveats
Our scores represent technical automation potential, not guaranteed job loss. As Anthropic's March 2026 study confirmed, actual AI adoption is just a fraction of theoretical capability. Actual displacement depends on:
- Economic feasibility of deploying AI vs. human labor
- Regulatory and legal frameworks in each country
- Social acceptance and trust in AI systems
- Speed of AI technology advancement
- New roles and industries created by AI (1.3M new AI jobs created 2024–2026 per LinkedIn)
- Wage premium in AI-exposed sectors (16.7% above national average per Fed Dallas)
Questions about our methodology?
We believe in transparency. If you think a score is wrong, we want to hear about it.
Check Your Job →