Implementation guide
Structure Incident Investigations
Detailed training workflow for Structure Incident Investigations in EHS & Safety.
ehsincident
Guided walkthrough
The Problem: Incident reports are often incomplete, blame-focused, and miss systemic root causes. Timeline Builder Input raw facts — AI structures a chronological event sequence. RCA Engine AI applies multiple root cause methods (Fishbone, Fault Tree) simultaneously. 5 Whys Analysis AI performs an iterative '5 Whys' review to find systemic process failures vs. human error.
Advanced implementation notes
Multi-Method Root Cause Analysis Engine Implement a comprehensive incident investigation framework that applies 5 Whys, Fishbone (Ishikawa), Fault Tree Analysis, and TapRooT® methodology simultaneously to surface systemic organizational failures — not just proximate causes. Fact Collection Input witness statements, photos, environmental data, and equipment logs. AI structures these into a verified timeline with 'confirmed', 'probable', and 'unverified' fact categories. Barrier Analysis AI identifies which safety barriers (engineering controls,
procedures, training, PPE) were present, absent, or failed. Maps to James Reason's Swiss Cheese Model. Multi-Method RCA AI simultaneously runs: 5 Whys (linear causation), Fishbone (6M categories: Man, Machine, Method, Material, Measurement, Mother Nature), and Fault Tree (Boolean logic gates). Systemic Pattern Detection Cross-reference this incident against the last 24 months of incidents in the Vault. AI identifies recurring themes (e.g., 'training gaps in contractors' appearing in 4 of last 7 incidents). CAPA Generation Generate Corrective and
Preventive Actions with: specific action description, responsible person, due date, effectiveness verification method, and leading indicator to track. Always separate 'proximate cause' from 'root cause' — AI should drill past 'operator error' to find why the operator was in a position to make that error. Include a 'Management System Failure' category in every Fishbone — leadership decisions are root causes in 70% of serious incidents. Assign each CAPA a leading indicator, not just a completion date — 'training completed' is lagging; 'correct procedure
observed in 3 field audits' is leading. Don't stop at '5 Whys' — the method is useful but linear. Complex incidents require Fault Tree's Boolean logic to capture multiple contributing factors. Don't use passive voice in findings ('the valve was left open') — identify the specific system failure that allowed it ('no post-maintenance verification checklist existed'). Don't assign CAPAs to 'All Employees' — vague ownership guarantees zero accountability. The 'Pre-Mortem' Technique Use the incident investigation prompt in reverse before high-risk activities
begin: 'Imagine this task resulted in a serious injury. What went wrong?' AI generates a predictive incident report that reveals gaps in your existing controls before anyone gets hurt.