17 KiB
17 KiB
Business Requirement Document
Centralized AI-Powered HSE Incident Management & Reporting System
Client: Setia Corporation — 3PL Warehouse Operations
Prepared by: Wee Ihan Yap
Version: 1.0
Date: 12 July 2026
Status: Final
1. Executive Summary
Setia Corporation currently manages Health, Safety and Environment (HSE) incident reporting through a combination of WhatsApp messages, Excel spreadsheets, and paper-based forms. This process creates significant gaps: delayed notifications, incomplete evidence, inconsistent classification, and manual workload for statutory DOSH reporting.
This document defines the business requirements for a centralized, AI-powered HSE Incident Management System (IMS) that replaces these manual processes. The system captures every incident digitally, runs a structured workflow from report through closure, tracks corrective actions, and generates statutory JKKP reports automatically — eliminating the administrative burden on HSE staff and ensuring zero missed regulatory deadlines.
2. Business Objectives
- Replace WhatsApp/Excel/paper incident reporting with a single digital system capturing all incident types.
- Enforce a structured, auditable workflow: report → triage → investigation → CAPA → verification → closure.
- Track all Corrective and Preventive Actions (CAPAs) to completion with automated escalation for overdue items.
- Provide management with a live dashboard replacing monthly manual report compilation.
- Ensure full compliance with Malaysian DOSH reporting obligations (NADOPOD 2004) without manual form-filling.
- Use AI (Claude API) to reduce HSE admin burden: drafting summaries, suggesting root causes, flagging incomplete reports, and detecting risk patterns.
3. Stakeholders
| Role | Representative(s) | Responsibility |
|---|---|---|
| Sponsor / Business Owner | Ms. Agnes, Mr. Terence | Approve requirements, budget, go-live sign-off |
| HSE Lead | Mr. Yap | Primary system user; investigation, CAPA, DOSH filing |
| Operations | Mr. Jensen | Supervisor workflow, CAPA ownership |
| Floor Staff / Reporters | All employees, guards | Incident submission via QR scan |
| System Developer | Wee Ihan Yap | Design, build, deployment |
4. User Roles & Access
| Role | Who | Access Level | Primary Responsibility |
|---|---|---|---|
| Reporter | Any employee, supervisor, witness, security guard | Submit reports via QR/app; view own reports only | Report incidents and hazards immediately |
| Operation Supervisor | Warehouse/shift supervisor | Full access to incidents in their site/zone | Secure area, first response, initial classification |
| HSE Officer | Safety Assistant / HSE team | Full access across all sites | Investigation, RCA, CAPA assignment, verification, closure, DOSH filing |
| CAPA Owner | Department head (Ops, Maintenance, HR, etc.) | CAPA items assigned to their department only | Complete corrective actions, upload proof |
| Management | Ms. Agnes, Mr. Terence, Mr. Yap, Mr. Jensen | Read-only dashboard across all sites | Review trends, approve budgets, audit readiness |
| System Admin | IT / appointed super-user | Full configuration access | User management, site/zone setup, form configuration |
Every action must be attributable to a logged-in user — no anonymous edits.
5. Scope
5.1 In Scope
- Digital incident capture for all seven incident types (see §6)
- Full incident lifecycle: report → triage → investigation → CAPA → verification → closure
- Evidence management: photos, videos, documents at every lifecycle stage
- CAPA board with auto-escalation and effectiveness re-check
- Email and WhatsApp Business notifications
- In-app notification bell and badge
- DOSH compliance: automated JKKP 6, JKKP 7, JKKP 8 PDF and CSV generation
- AI-assisted features: quality check, severity suggestion, similar incident retrieval, RCA/CAPA drafting, risk heatmap
- Multi-site, multi-zone QR-based reporting
- Multi-language UI (English, Bahasa Malaysia, Mandarin)
- Mobile-first, offline capture with auto-sync on reconnect
- Dashboard: leading/lagging indicators, site/zone heatmap, CAPA on-time rate, trend charts
- Admin panel: user management, site/zone CRUD, role assignment
- Full audit trail (every action logged with user + timestamp)
- 5-year data retention per DOSH requirements
5.2 Out of Scope
- WMS (Warehouse Management System) integration (optional future phase)
- Tamil language support (deferred unless workforce requires it)
- Third-party EHS platform integration
- External audit portal access
6. Incident Types
The system must support all seven incident types, each with a type-specific intake form:
- Injury / Medical Treatment Case — LTI and non-LTI
- Near Miss — fast, low-friction form (near-miss volume is the primary leading safety indicator)
- Unsafe Condition / Hazard Observation — proactive, not tied to an event
- Property / Asset / MHE Damage — forklift, racking, dock equipment
- Environmental Incident — spill, leak, chemical release, waste
- Security Incident — theft, unauthorized access
- Fire / Emergency Incident
7. Incident Reference Format
Every incident auto-generates a unique reference number:
SITE-YYYYMM-####
Example: KL01-202607-0042
8. Functional Requirements
8.1 Incident Reporting
- Reporter scans a site/zone-specific QR code or opens the app
- System pre-fills site and zone from QR token (no manual entry)
- Reporter selects incident type and completes type-specific intake form
- Minimum one photo required for injury reports; video and documents optional
- System auto-generates incident reference number and timestamps report
- Automatic notification fires immediately to the relevant Supervisor and HSE Officer
- Offline capture supported: form data queued in browser (IndexedDB) and synced on reconnect
8.2 Triage & Initial Response
- Supervisor or HSE Officer confirms or reclassifies incident type
- Assigns severity level (1–5); AI may suggest a level but a human always confirms
- Workflow branches by type:
- Injury: medical/first-aid path, LTI/non-LTI classification, lost-day tracking
- Asset/MHE: emergency shutdown/LOTO, operator-error check, HR/discipline path if applicable
- Environmental: containment steps, spill-kit deployment, environmental authority check
- Near miss / hazard: skip to root-cause and CAPA directly
- System automatically evaluates incident data against NADOPOD 2004 rules and presents the applicable obligation (immediate DOSH notification / JKKP 6 / JKKP 7 / JKKP 8) as a checklist for the HSE Officer
8.3 Investigation & CAPA
- HSE enters witness statements, alcohol/urine test result (if applicable), evidence
- Structured root-cause analysis via selectable template: 5-Why or Fishbone (not free text only)
- Every CAPA item must record: description, responsible department/owner, due date, priority, and root-cause linkage
- Auto-escalation ladder:
- 3 days before due date: reminder to owner
- On due date: notify owner
- 3 days overdue: notify owner's manager
- 7 days overdue: notify HSE Officer, flag red on dashboard
8.4 Verification & Closure
- CAPA owner uploads completion evidence (photo/document proof) before marking CAPA as done
- HSE verifies effectiveness; if not effective, CAPA reopens (not closed with open gap)
- Once all CAPAs verified, HSE closes incident
- On closure: record locks against further edits; only addenda can be appended
- Closed incident automatically enters JKKP 8 annual register
8.5 Evidence Management
Evidence must be attachable at every lifecycle stage, not only at initial report:
| Stage | Expected Evidence |
|---|---|
| Report | Scene photo/video, hazard photo |
| Response | LOTO tag photo, first-aid record, medical referral letter |
| Investigation | Witness statement scans, CCTV export, equipment inspection report, alcohol/urine test result |
| CAPA | Before/after photos, purchase receipts, training attendance sheets, updated SOP |
| Verification | Final confirmation photo/video that corrective action is in place and effective |
Requirements:
- Accepted formats: JPG, PNG, HEIC, MP4, MOV, PDF, DOCX, XLSX
- Max file size: configurable (recommended 200 MB for video; compress on upload)
- Every file records: uploader, timestamp, incident ID, stage, immutable SHA-256 file hash
- Files retained minimum 5 years; never auto-deleted
- Thumbnail/preview generation so HSE can review without downloading
8.6 Notifications
| Channel | Use |
|---|---|
| Formal records: investigation assignment, CAPA assignment, closure notifications | |
| WhatsApp Business API | Time-critical alerts: new serious incident, CAPA overdue escalation |
| In-app notification bell | All events for all users; unread badge count; dropdown list |
8.7 Dashboard & Analytics
- Total incidents, near misses, severity rate, open vs. closed counts
- Leading vs. lagging indicator split (near miss/hazard = leading; injury/LTI = lagging)
- Site/zone/shift heatmap — critical for multi-warehouse operations
- CAPA on-time completion rate (%)
- Top incident category and top root cause, trended over 12 months
- DOSH-reportable incident count and filing status (filed / pending / overdue)
- AI rising-risk zone flags: zones with statistically increasing incident frequency
- Export to PDF/Excel for board reporting
8.8 AI-Assisted Capabilities
All AI outputs are suggestions that a human reviews and approves — never auto-submitted to DOSH and never auto-closed without human sign-off.
| Capability | Description | Business Value |
|---|---|---|
| Report quality check | Flags incomplete reports before submission (missing photo on injury report, vague description) | Fixes delayed/incomplete information problem |
| Severity/category suggestion | Suggests severity level (1–5) and incident category from free-text description | Speeds triage, reduces classification inconsistency |
| Similar incident retrieval | Surfaces top-5 past incidents with similar description/location/equipment via vector similarity | Reveals recurring hazards; supports trend detection |
| RCA/CAPA drafting assistant | Suggests likely root causes and draft corrective actions from investigation notes | Cuts write-up time, improves CAPA consistency |
| JKKP form auto-fill | Generates JKKP 6/7 PDF drafts and JKKP 8 annual register from stored data | Removes single biggest admin burden |
| Risk heatmap / prediction | Combines near-miss, incident, and hazard data by site/zone/shift to flag rising-risk areas | Predictive safety capability tuned to Setia's own warehouses |
All AI suggestions logged to audit trail: what was suggested and what the human ultimately chose.
8.9 DOSH Compliance (NADOPOD 2004)
The system encodes Malaysia-specific statutory reporting rules and automatically determines the applicable obligation:
| Situation | Obligation |
|---|---|
| Fatality or serious bodily injury (NADOPOD First Schedule: fracture, amputation, loss of sight) | Notify nearest DOSH office immediately; submit JKKP 6 within 7 days |
| Dangerous occurrence (Second Schedule: boiler explosion, structural collapse) regardless of injury | Notify DOSH immediately; submit JKKP 6 within 7 days |
| Other injury causing incapacity for more than 4 consecutive days | Submit JKKP 6 within 7 days |
| Occupational poisoning or disease (Third Schedule) | Submit JKKP 7 within 7 days |
| Any of the above | Also logged in JKKP 8 annual register; retained on-site 5 years; submitted to DOSH before 31 January each year |
8.10 Admin Management
- User management: invite by email, assign role and site, activate/deactivate
- Site CRUD: name, address, region, active flag
- Zone CRUD: name per site, QR code generation and download per zone
- All admin actions logged to audit trail
9. Non-Functional Requirements
| Category | Requirement |
|---|---|
| Mobile-first | Must work on low-end Android phones common on warehouse floors |
| Offline | Incident form submittable offline; data queues locally (IndexedDB) and auto-syncs on reconnect |
| Multi-language | English, Bahasa Malaysia, Mandarin (all three available at all times via language switcher) |
| Multi-site | Site and zone are first-class fields on every record from day one |
| Access control | Role-based access enforced at the database level (Supabase RLS), not only in UI |
| Audit trail | Every create/edit/status-change/file-upload logged with user and timestamp; immutable |
| Data retention | Minimum 5 years per DOSH JKKP 8 requirement; evidence files never hard-deleted |
| Performance | Incident list uses server-side pagination; no unbounded queries |
| Security | API keys server-side only; never exposed to client; RLS on all tables |
10. System Architecture Summary
Stack:
| Layer | Technology |
|---|---|
| Frontend + API routes | Next.js 15 (App Router) |
| Database + Auth + Storage | Supabase (Postgres + RLS + pgvector) |
| Hosting | Vercel (frontend) + Supabase cloud |
| AI | Claude API (Anthropic) — server-side only |
| Resend | |
| Meta WhatsApp Business Cloud API | |
| PDF generation | pdf-lib (JKKP 6/7 form fill) |
| QR codes | qrcode npm package |
Deployment: Vercel (frontend) + Supabase cloud. Custom domain hse.setiacorp.com once MVP validated. All secrets stored as environment variables — never committed to code.
11. Key Database Entities
| Entity | Purpose |
|---|---|
| sites | Warehouse locations |
| zones | Named areas within a site, each with a unique QR token |
| users | All system users with role, department, site assignment |
| incidents | Core record: type, site, zone, severity, status, lifecycle timestamps |
| evidence_files | Files attached at each lifecycle stage with immutable hash |
| investigations | RCA method, findings, root cause summary, test results |
| capa_actions | CAPA items with owner, due date, priority, status, effectiveness recheck date |
| dosh_reports | JKKP 6/7/8 records with filing status and generated PDF link |
| notifications_log | All notifications sent across all channels |
| audit_log | Immutable record of every system action |
12. Development Phases
| Phase | Scope | Status |
|---|---|---|
| 0 | Foundation: scaffold, auth, DB migrations, QR codes | Complete |
| 1 | Core reporting: incident form, inbox, evidence upload, email notification | Complete |
| 2 | Investigation + CAPA: triage, 5-Why/fishbone, CAPA board, verification, JKKP 6/7 PDF | Complete |
| 3 | AI features + dashboards: Claude integration, pgvector similar incidents, role dashboards, CSV export | Complete |
| 4 | Scale & polish: WhatsApp notifications, i18n EN/MS/ZH, PWA offline capture, CAPA effectiveness recheck | Complete |
| 5 | Usability & compliance: notification bell, closure lock, pagination, type-specific intake, witness/alcohol UI, JKKP 8, admin management, evidence thumbnails | Complete |
| 6 | Analytics & predictive: 12-month trend chart, top root causes trended, AI rising-risk zone flags | Complete |
13. Success Metrics
| Metric | Target |
|---|---|
| Digital incident capture rate | 100% within 30 days of launch (zero WhatsApp-only reports) |
| CAPA on-time closure rate | Above 85% within 3 months |
| Incident-to-HSE notification time | Under 2 minutes (vs. current WhatsApp-dependent manual forwarding) |
| Monthly HSE report preparation time | Reduced from days to minutes (auto-generated from dashboard) |
| Missed DOSH statutory reporting deadlines | Zero |
14. Assumptions & Constraints
- All warehouse sites have internet access sufficient for mobile web browsing; offline mode covers low-signal periods only.
- Users are expected to have a smartphone (Android minimum); desktop access available for HSE Officers and management.
- Supabase free/starter tier sufficient for initial rollout; video-heavy storage may require upgrade to Supabase Pro or Cloudflare R2 migration at scale.
SUPABASE_SERVICE_ROLE_KEYmust be set in production environment for admin user-invite functionality.- WhatsApp Business API requires approved Meta Business account and message templates before production use.
- VPS cron jobs must be registered for CAPA escalation and effectiveness recheck automation post-deployment.
- Evidence files must never be deleted; storage cost is a known ongoing operational expense.
15. Glossary
| Term | Definition |
|---|---|
| CAPA | Corrective and Preventive Action |
| DOSH | Department of Occupational Safety and Health (Malaysia) |
| HSE | Health, Safety and Environment |
| JKKP 6 | Malaysia statutory form — notification of accident/dangerous occurrence |
| JKKP 7 | Malaysia statutory form — notification of occupational poisoning/disease |
| JKKP 8 | Malaysia statutory form — annual register of accidents/occupational diseases |
| LTI | Lost Time Injury (injury resulting in at least one day away from work) |
| LOTO | Lockout/Tagout (energy isolation safety procedure) |
| MHE | Material Handling Equipment (forklifts, pallet jacks, etc.) |
| NADOPOD 2004 | Notification of Accident, Dangerous Occurrence, Occupational Poisoning and Occupational Disease Regulations 2004 |
| pgvector | Postgres extension for vector similarity search (used for similar incident retrieval) |
| QR | Quick Response code (used for site/zone identification on incident report form) |
| RCA | Root Cause Analysis |
| RLS | Row-Level Security (Supabase/Postgres feature enforcing data access at DB level) |
| 3PL | Third-Party Logistics |