Skip to content

DEMO_GUIDE.md

Latest commit

 

History

History
218 lines (173 loc) · 9.86 KB

DEMO_GUIDE.md

File metadata and controls

218 lines (173 loc) · 9.86 KB
Raw

Demo Guide — Claim Processing Services

End-to-End Claim Lifecycle Demo

This guide walks through the complete claim journey from filing to settlement, demonstrating all three personas.

Platform URL: https://.cloudfront.net

Test Accounts

Persona Login URL User ID Password
Policyholder /login/policyholder testuser2 REDACTED-TEST-VALUE
Claims Adjudicator /login/adjudicator adjudicator1 REDACTED-TEST-VALUE
System Admin /login/admin admin1 Admin12345

Claim Lifecycle Flow

┌──────────┐    ┌───────────┐    ┌──────────────┐    ┌──────────┐    ┌────────────┐    ┌────────┐
│  DRAFT   │───▶│ SUBMITTED │───▶│ UNDER REVIEW │───▶│ APPROVED │───▶│ SETTLEMENT │───▶│ CLOSED │
└──────────┘    └───────────┘    └──────────────┘    └──────────┘    └────────────┘    └────────┘
                                         │                                                    ▲
                                         │           ┌──────────┐                             │
                                         └──────────▶│ REJECTED │─────────────────────────────┘
                                                     └──────────┘
Status Who Acts Action
Draft Policyholder Creates claim, fills details
Submitted/Under Review System Auto-transitions on submit
Approved Adjudicator Reviews and approves with settlement amount
Rejected Adjudicator Reviews and rejects with reason
Settlement Adjudicator Initiates payment processing
Closed Adjudicator Finalizes the claim

Demo Script (Step by Step)

Act 1: Policyholder Files a Claim

  1. Open the platform: https://.cloudfront.net
  2. Show the Landing Page — Three persona cards (Policyholder, Adjudicator, Admin)
  3. Click "Policyholder" → Blue-themed login page
  4. Login with testuser2 / REDACTED-TEST-VALUE
  5. Dashboard shows "My Claims" with stats and recent claims
  6. Click "File New Claim" button
  7. Step 1 - Select Type: Choose "Health Insurance" 🏥
  8. Step 2 - Details:
    • Amount: 75000 (₹75,000)
    • Policy Number: POL-HEALTH-2024-001
    • Incident Date: Select today's date
    • Description: Hospitalization for acute appendicitis surgery at Apollo Hospital, Mumbai. 3-day stay including surgery, post-operative care, and medications.
  9. Step 3 - Documents: Select a file (optional) or skip
  10. Step 4 - Review: Verify all details, click "Submit Claim"
  11. Result: Claim is created and submitted → Status moves to "Under Review"
  12. Navigate to Claims list — See the new claim with "Under Review" status
  13. Click the claim — See the Claim Journey timeline showing progress
  14. Logout (top-right corner)

Act 2: Adjudicator Reviews and Approves

  1. Go back to Landing Page: https://.cloudfront.net
  2. Click "Claims Adjudicator" → Green-themed login page
  3. Login with adjudicator1 / REDACTED-TEST-VALUE
  4. Dashboard shows "Claims Queue" with all claims including the new one
  5. Click the health claim just submitted
  6. Show the Claim Journey — Currently at "Under Review" step
  7. Show action buttons: "Approve" and "Reject" are available
  8. Click "Approve" → Modal appears
  9. Enter Settlement Amount: 70000 (₹70,000 — slightly less than claimed)
  10. Add Notes: Approved after verification. Settlement amount adjusted per policy terms.
  11. Click "Confirm Approval"
  12. Result: Status moves to "Approved" → Timeline updates
  13. Click "Initiate Settlement" button (now visible)
  14. Result: Status moves to "Settlement"
  15. Click "Close Claim" button
  16. Result: Status moves to "Closed" → Full lifecycle complete!
  17. Show the Claim History section — All status changes with timestamps

Act 3: Admin Monitors the System

  1. Go back to Landing Page
  2. Click "System Admin" → Purple-themed login page
  3. Login with admin1 / Admin12345
  4. Dashboard shows system overview with stats
  5. Navigate to "Analytics" — Show charts with real data:
    • Claims by Type (bar chart)
    • Status Distribution (donut chart)
    • Monthly Trend (vertical bars)
    • Performance table
  6. Navigate to "Users" — Show user management:
    • All registered users with roles
    • Filter by role
    • "Add User" button (creates new users)
    • "Edit" button (change roles)
    • "Deactivate" button
  7. Navigate to "Claims" — Admin can see all claims across all users

Act 4: Rejection Flow (Optional)

  1. As Policyholder: File another claim (e.g., Motor Insurance, ₹2,00,000)
  2. As Adjudicator: Open the claim → Click "Reject"
  3. Enter reason: Claim amount exceeds policy coverage limit. Please resubmit with correct documentation.
  4. Click "Confirm Rejection"
  5. Result: Status moves to "Rejected"
  6. Click "Close Claim" → Status moves to "Closed"
  7. As Policyholder: Login and see the rejected claim with reason displayed

Key Demo Talking Points

Business Value

  • "80% of straightforward claims settled within 24 hours"
  • "70% automation rate — minimal manual intervention"
  • "Real-time visibility for policyholders — no more calling support"

Technical Highlights

  • "7 microservices on AWS ECS Fargate — independently scalable"
  • "React TypeScript frontend on CloudFront — global CDN delivery"
  • "DynamoDB for operational data — millisecond response times"
  • "Role-based access control — each persona sees only what they need"
  • "Full audit trail — every status change is logged with timestamp"

Architecture

  • "Python FastAPI microservices — modern, async, high-performance"
  • "AWS CDK Infrastructure as Code — 13 stacks, fully reproducible"
  • "Service discovery via AWS Cloud Map — no hardcoded URLs"
  • "Designed for AI integration — Amazon Bedrock for fraud detection (ready to enable)"

Troubleshooting During Demo

Issue Quick Fix
Login fails Check credentials above, try registering a new user
Claim submit shows error Ensure description is 10+ characters
Adjudicator sees no claims Claims must be submitted (not just draft)
Page shows blank after action Hard refresh (Cmd+Shift+R)
API timeout Wait 5 seconds and retry — ECS cold start
Analytics "UndefinedTable" error Run: curl -X POST .../api/analytics/sync -H "Authorization: Bearer $TOKEN"
CloudFront returns HTML for API Re-add /api/* cache behavior pointing to alb-api origin
Notification "AccessDenied" on Scan Already fixed — IAM inline policy added to task role
Datadog shows service:cloudwatch Log group tags set — wait for new logs (old ones keep old tag)

Post-Demo Q&A Answers

Q: How does fraud detection work? A: Amazon Bedrock (Claude Sonnet 4.5) analyzes claim patterns, document consistency, and historical data. Currently in demo mode — ready to enable with Bedrock model access.

Q: How does document OCR work? A: Amazon Textract extracts text, Amazon Comprehend identifies entities (names, dates, amounts). Auto-populates claim fields. Document service is deployed and connected.

Q: What about notifications? A: Email (SES) and SMS (SNS) notifications are configured. Notification service is deployed — requires SES domain verification for production use.

Q: Can this scale? A: ECS Fargate auto-scales 1→3 tasks per service. DynamoDB on-demand scales automatically. Designed for 1,000+ claims/day at current config, easily scalable to 100K+.

Act 5: DevOps Agent Demo (Stress Testing)

Setup

cd scripts/stress-testing
pip install boto3 requests

Show Healthy State

  1. Login as Admin → Dashboard shows all services 🟢 Healthy
  2. Open CloudWatch Dashboard: https://console.aws.amazon.com/cloudwatch/home?region=us-east-1#dashboards:name=ClaimFlow-Operations
  3. Show all metrics are flat/normal
  4. Show CloudWatch Alarms: https://console.aws.amazon.com/cloudwatch/home?region=us-east-1#alarmsV2: — all 🟢 OK

Inject Stress

# Option 1: API flood (causes CPU spikes)
python stress_api.py --profile my-profile --duration 120 --rps 50

# Option 2: Kill services (causes outages)
python stress_ecs.py --profile my-profile --action scale-down

# Option 3: Combined chaos
python stress_all.py --profile my-profile --level medium --duration 180

Observe Degradation

  1. Admin Dashboard → Services go 🔴 Unhealthy, response times spike
  2. CloudWatch Dashboard → CPU/Memory graphs spike, error counts increase
  3. CloudWatch Alarms → Multiple alarms fire 🔴 (HighCPU, 5xxErrors, UnhealthyTargets)
  4. DevOps Agent identifies: "ECS CPU at 95%, 3 services unhealthy, DynamoDB throttling detected"

Recovery

python recover_all.py --profile my-profile

Verify Recovery

  1. Admin Dashboard → All services return to 🟢 Healthy
  2. CloudWatch Alarms → Return to 🟢 OK
  3. DevOps Agent confirms: "All services recovered, metrics normalized"

Key Talking Points

  • "42 CloudWatch alarms monitoring every layer of the stack"
  • "Automated detection of CPU spikes, memory pressure, throttling, and service outages"
  • "DevOps Agent can identify root cause across multiple services simultaneously"
  • "One-command recovery restores full platform health"