Key Takeaways
- Transform passive agent-driven model into proactive, SLA-bound field service management
- Central booking system removes agent-initiated customer contact chaos
- Auto-assignment algorithm uses weighted scoring: proximity, availability, performance, workload
- Multi-tier escalation (L0-L3) ensures no lead falls through the cracks
- Target: 60-minute TAT from lead assignment to KYC completion
EXECUTIVE SUMMARY
This document presents a comprehensive system design for transforming a passive, agent-driven KYC verification model into a proactive, SLA-bound field service management system.
The Problem
Many KYC service providers offer field verification services to partner banks through a network of gig agents. The current system operates on a passive model with several critical inefficiencies:
Voluntary Lead Acceptance
Agents view available leads in their pincode and choose whether to accept them. Many leads remain unprocessed.
No Time Accountability
No TAT enforcement means agents take arbitrary time to complete KYC, causing customer dissatisfaction.
Decentralized Scheduling
Agents independently contact customers and schedule appointments, leading to coordination chaos.
Zero Visibility
Banks have no real-time visibility into lead status, agent location, or estimated completion times.
Business Impact
- Bank partners facing regulatory penalties due to KYC delays
- Customer churn from poor experience and long wait times
- Revenue leakage from unprocessed leads
- Agent attrition due to inconsistent earnings
Current vs. New Flow
Current Flow (As-Is)
New Flow (To-Be)
System Architecture
The proposed system follows a microservices architecture with event-driven communication, enabling real-time state management and horizontal scalability.
External Layer
API Gateway & Load Balancer
Microservices Layer
Event Bus
Apache Kafka / RabbitMQData Layer
Backend
- • Node.js / Go microservices
- • GraphQL for mobile apps
- • REST for bank integrations
- • gRPC for inter-service
Infrastructure
- • Kubernetes on AWS EKS
- • CloudWatch monitoring
- • S3 for document storage
- • CloudFront CDN
Real-time
- • WebSocket for live tracking
- • Firebase for push
- • Redis pub/sub for events
- • H3 geospatial indexing
Auto-Assignment Algorithm
The auto-assignment engine uses a weighted scoring algorithm inspired by ride-hailing dispatch systems (Uber, Ola) but optimized for scheduled appointments.
Score = (0.35 × Proximity) + (0.25 × Availability) + (0.25 × Performance) + (0.15 × Workload)Proximity Score (W₁ = 0.35)
1 - (distance / max_distance)
Euclidean distance using H3 geohashing
Availability Score (W₂ = 0.25)
slots_free / total_slots
Current task load vs. maximum capacity
Performance Score (W₃ = 0.25)
(completion_rate × 0.5) + (rating × 0.5)
Historical KYC completion and customer ratings
Workload Balancing (W₄ = 0.15)
1 - (daily_tasks / avg_daily_tasks)
Ensures fair distribution across pool
Fallback Mechanisms
- 1. Expand search radius by 2km increments
- 2. Include agents from adjacent pincodes
- 3. Offer later time slot to customer
- 4. Manual assignment by operations team
- 5. Video KYC as last resort alternative
SLA & Escalation Matrix
60-Minute SLA Breakdown
| Stage | Time Window | SLA |
|---|---|---|
| Customer Contact & Slot Selection | 0-5 min | 5 min |
| Agent Assignment | 5-10 min | 5 min |
| Agent Acknowledgment | 10-12 min | 2 min |
| Travel to Customer | 12-35 min | 23 min |
| KYC Verification | 35-55 min | 20 min |
| Document Upload & Closure | 55-60 min | 5 min |
Escalation Matrix
L0 - Warning
70% of stage SLASystem nudge to agent • Notified: Agent only
L1 - Alert
Stage SLA breachedSupervisor notification • Notified: Agent + Team Lead
L2 - Critical
Overall SLA 80% elapsedReassignment initiated • Notified: Operations Manager
L3 - Breach
60-min SLA breachedIncident logged + penalty • Notified: Regional Head + Bank
Edge Case Handling
Agent accepts but doesn't go
Detection: No GPS movement for 10 minutes
- 1. System sends warning at 5 min
- 2. Auto-reassign to backup agent at 10 min
- 3. Original agent receives penalty
Customer not available at scheduled time
Detection: Agent marks "Customer Unavailable"
- 1. Geofence verification ensures agent is at location
- 2. Agent makes 2 call attempts with 5 min gap
- 3. Lead goes to rebooking queue
Customer refuses KYC
Detection: Agent selects "Customer Refused"
- 1. Mandatory reason selection required
- 2. Lead returned to bank with refusal reason
- 3. Agent receives partial payment for visit
Agent goes offline mid-task
Detection: No heartbeat/GPS for 3 minutes
- 1. Immediate escalation to supervisor
- 2. Backup agent dispatched if no response
- 3. Customer proactively informed of delay
No agents available in pincode
Detection: Assignment engine returns empty list
- 1. Expand search to neighboring pincodes
- 2. Offer Video KYC alternative
- 3. Alert operations to deploy agents to area
Admin Dashboard Design
Lead Management
- • Lead pipeline with stage filters
- • Bulk actions (reassign, cancel)
- • Search by lead ID, customer phone
Agent Management
- • Agent roster with status
- • Performance scorecards
- • Penalty and incentive tracker
Analytics
- • Bank-wise SLA compliance
- • Pincode heat maps
- • Forecasting and capacity
Agent App Design
Home Dashboard
- • Today's assigned tasks
- • Earnings tracker
- • Go Online/Offline toggle
- • Quick stats (completed, pending)
Task Detail
- • Customer name and address
- • KYC type and documents required
- • Appointment time with countdown
- • Call customer / Navigate buttons
KYC Capture
- • Document scanner with AI assist
- • Face match verification
- • E-signature capture
- • Submit for review
Task Status Flow
Implementation Roadmap
Phase 1: Foundation
Weeks 1-4Core microservices, database schema, basic admin dashboard, agent app MVP, bank API integration
Phase 2: Intelligence
Weeks 5-8Auto-assignment algorithm, H3 geospatial indexing, real-time GPS tracking, SLA monitoring
Phase 3: Automation
Weeks 9-12Automated escalation workflows, agent performance scoring, incentive/penalty automation
Phase 4: Optimization
Weeks 13-16ML-based demand prediction, dynamic slot capacity, agent route optimization, Video KYC fallback
Success Metrics
THE BOTTOM LINE
The proposed KYC Field Verification System transforms any passive lead marketplace into an active, SLA-driven service delivery platform.
By centralizing appointment booking, implementing intelligent auto-assignment, and enforcing strict escalation protocols, the 60-minute TAT target becomes achievable and measurable. This system architecture provides the foundation for scaling to additional banks and potentially other field verification services beyond KYC.
Kaiross Team
System Architecture & Product Strategy