Methodology
A detailed look at how this tool assesses workforce automation exposure and skill implications.
Overview
The Workforce Task Intelligence methodology provides a structured approach to understanding how AI capabilities may impact specific job roles. Rather than making broad predictions about job displacement, the tool focuses on task-level analysis to provide actionable insights.
The approach is grounded in the ILO Working Paper 140 (2025) framework for assessing generative AI exposure, combined with comprehensive occupational data from the U.S. Department of Labor's O*NET database.
This approach recognizes that most jobs consist of a portfolio of tasks with varying degrees of automation potential. The same role may have some tasks that are highly automatable while others remain firmly in the human domain.
Taxonomy Resolution
What the Tool Does
The tool maps input job titles to standardized occupational classifications using the O*NET (Occupational Information Network) database. A fuzzy search algorithm matches against 57,521 job titles (primary titles plus alternate titles) to find the best match.
Data Sources
- O*NET Database 30.1 (December 2025) - 1,016 occupations
- Bureau of Labor Statistics Standard Occupational Classification (SOC)
- 57,521 searchable job titles (primary + alternates)
Limitations
- O*NET is US-centric; international job titles may not match well
- Emerging roles may not have established classifications
- Organizational variations in job definitions not captured
Task Decomposition
What the Tool Does
Each role is broken into its constituent tasks using O*NET task statements. The tool analyzes up to 25 tasks per occupation, capturing 77% of occupations completely. The database contains 18,796 total task statements across all occupations.
Task Classification Framework
Based on ILO Working Paper 140, each task is classified into one of three categories:
Tasks where AI can perform the core function with minimal human oversight
Tasks where AI enhances human capability but human judgment remains essential
Tasks that remain primarily human due to physical, interpersonal, or judgment requirements
Limitations
- Occupations with more than 25 tasks are partially analyzed
- Informal tasks and organizational context not fully captured
- Task interdependencies may affect automation feasibility
Six-Dimension Assessment
What the Tool Does
Each task is evaluated using the ILO's six-dimensional assessment framework. Starting from a baseline score of 50, adjustments are made based on each dimension to arrive at an automation potential score (0-100).
Assessment Dimensions
1. Task Structure
Structured, rule-based tasks score higher; unstructured tasks score lower
2. Cognitive vs Physical
Pure information tasks score higher; physical/hands-on tasks score lower
3. Routine vs Novel
Repetitive tasks score higher; unprecedented situations score lower
4. Human Judgment Requirement
Objective criteria score higher; subjective judgment scores lower
5. Interpersonal Intensity
Solo tasks score higher; relationship-dependent tasks score lower
6. Stakes & Accountability
Low-stakes tasks score higher; high-stakes decisions score lower
Capability Level Scenarios
Classification thresholds adjust based on the selected AI capability assumption:
Higher thresholds for automation (75+), more tasks classified as Retain
Standard thresholds (70+/30+), balanced assessment of current capabilities
Lower thresholds for automation (65+), assumes rapid capability advancement
Limitations
- AI capabilities evolving rapidly; assessments reflect analysis date
- Breakthrough capabilities may not follow historical trends
- Industry-specific AI adoption rates vary significantly
Exposure Calculation
What the Tool Does
Task-level automation potential is aggregated into an overall exposure score. The distribution shows what percentage of tasks fall into each category (Automate, Augment, Retain), providing a clear picture of how the role will transform.
Exposure Categories
Limitations
- Does not account for industry-specific adoption barriers
- Regulatory constraints may significantly delay automation
- Technical potential differs from organizational readiness
Skills Inference
What the Tool Does
Based on the task classifications, the tool infers skill implications across three categories to provide actionable workforce development guidance:
Skills associated with automatable tasks that will decrease in value
Skills that need to transform for human-AI collaboration (highest training priority)
Uniquely human skills that become more valuable as AI handles routine work
Limitations
- Skills are inferred from task analysis, not validated against skill databases
- Individual development paths depend on current competencies
- Organizational context affects skill prioritization
Data Sources & References
- O*NET Database 30.1:December 2025 release. 1,016 occupations, 18,796 task statements, 57,521 job titles.onetcenter.org
- ILO Working Paper 140:"Generative AI and Jobs: A Refined Global Index of Occupational Exposure" (2025). Six-dimensional assessment framework.ilo.org
- Claude AI (Sonnet):Anthropic's Claude claude-sonnet-4-20250514 model performs task classification and reasoning.anthropic.com
- Bureau of Labor Statistics:Standard Occupational Classification (SOC) system for occupation taxonomy
Technical Implementation
Analysis Pipeline
- Fuzzy search matches job title to O*NET occupation (~100ms)
- Retrieve task statements from O*NET database (~100ms)
- Send tasks to Claude API with ILO framework prompt (60-90s)
- Parse structured JSON response with classifications and reasoning
- Calculate exposure statistics and infer skill implications
Streaming Response
Results stream progressively via Server-Sent Events (SSE): O*NET match appears in ~1 second, followed by task list, then full classification results as AI analysis completes.
Cost & Performance
Each analysis costs approximately $0.05-0.06 in API usage and takes 60-90 seconds to complete. Results are not cached, ensuring fresh analysis each time.
Important Disclaimer
This tool provides AI-generated analysis based on established occupational data and research frameworks. While it uses real O*NET data and ILO methodology, the classifications represent technical automation potential, not predictions of actual job changes.
Real-world workforce impact depends on organizational context, industry adoption rates, regulatory factors, economic considerations, and change management capabilities that vary significantly across employers and regions.