Portfolio Case Study · BFSI · Agentic AI

AI-Assisted KYC &
Customer Onboarding

An end-to-end product case study for an agentic AI onboarding platform — designed with explainability, audit, and human-in-the-loop as first-class requirements, not afterthoughts.

StatusPortfolio Case Study DomainBFSI · Fintech ComplianceFATF · EU AI Act · GDPR PatternAgentic · RAG · LLM
⚠️ Portfolio research project — design artifacts only. No client data, no production code, no client-specific claims. Targets are design intent, not measured production results.

The Problem

Banks lose customers in onboarding. Manual KYC reviews take days. AML screening produces ~30% false positives. Existing automation stops at OCR — judgment work still routes to humans without context, audit trails, or explainability.

This case study designs an agentic AI onboarding platform that assists humans on judgment work, while keeping every decision auditable, explainable, and reversible.

Business Impact — Design Targets

KPIIndustry BaselineDesign TargetDriver
Onboarding time2 days< 15 minParallel agents, OCR-first capture
Manual review rate100%< 20%Confidence-gated auto-approval
AML false positives~30%< 10%Multi-signal risk scoring
Customer drop-off~40%< 15%Async re-entry, clear UX
Audit evidence assemblyhours< 60 secEvent-sourced audit log

Architecture

Seven bounded-scope agents. Deterministic stage gates. Human-in-the-loop above the risk threshold. Full audit trail per event.

Agent 1

Intake

Session context, account type, jurisdiction routing.

Agent 2

Document Intelligence

OCR + structured extraction with per-field confidence.

Agent 3

KYC Validation

Jurisdiction rules, completeness, format checks.

Agent 4

Screening

Sanctions, PEP, adverse media via approved providers.

Agent 5

AML Risk Scoring

Multi-signal score 0–100 with factor breakdown.

Agent 6

Review Routing

Policy-driven routing to reviewer queue.

Agent 7

Audit

Event-sourced log; case evidence in < 60 sec.

Key Product Decisions

DecisionAlternativesWhy This Choice
Multi-agent architectureMonolithic LLMPer-step auditability; bounded failure
Human review for high-riskFull automationRegulatory mandate (FATF · EU AI Act Art. 14)
Risk score 0–100Binary approve/rejectExplainability; reviewer prioritization
Event-sourced audit logPeriodic snapshotsReconstruct any decision on demand
Fail-closed on uncertaintyFail-open with monitoringCompliance non-negotiable

Compliance Framework

Designed against named regulations — not generic "compliance" hand-waving.

FATF Recommendations

Rec. 10 CDD · Rec. 11 record-keeping · Rec. 12 PEP · Rec. 24 beneficial ownership.

EU AI Act

High-risk classification (Annex III §5). Art. 14 human oversight is a mandate, not a UX choice.

GDPR

Art. 22 right to human review · Art. 25 privacy by design · Art. 32 security.

BSA / CIP (US)

Identity verification within reasonable time · OFAC SDN screening · 5-yr retention.

Basel III CDD

Beneficial ownership ≥ 25%; ongoing CDD.

AMLD6 (EU)

Predicate offences; criminal liability for AML failures. Reviewer override always justified.

About the Author

Vittobha Vignesh S — Product Manager and Senior Product Owner. 10 years across BFSI, Healthcare, Telecom, and Enterprise Consulting. Scaled platforms to 20K enterprise users and 1M+ consumers across 20 countries.