Skip to main content
2019STERLING BANK

AI-Powered Financial Crime Prevention

Real-Time Fraud Detection Dashboard. Predictive monitoring that cut fraud by 34% and false positives by 47%, saving ₦155M.

Product DesignAIUX Engineering
33% reduction in inquiries€3.2M annual savings28% satisfaction improvement
Security and Fraud Detection

Overview

SterlingBankfacedmountingfinanciallossesfromsophisticatedfraudattacks.Theirreactiveapproachmeantfraudwasonlyinvestigatedaftercustomerfundswerecompromised,resultinginsignificantlegalcosts,customerdissatisfaction,andregulatoryscrutiny.

IdesignedanddevelopedanAI-poweredfrauddetectionplatformthatmonitorscustomerbehaviorpatternsinreal-time,transformingthebank'sapproachfromreactiveinvestigationtoproactiveprevention—allwhilemaintainingaseamlesscustomerexperience.

78%
Fraud Reduction
€2.3M
Annual Savings
92%
Satisfaction
6 Mo
Timeline

Roles & Responsibilities

Role

Solo UX Engineer

Research, Design, Frontend Development

Team

Cross-functional (5)

Data analytics, fraud specialists, engineers

Market

Nigeria

Fintech, Banking Security

The Challenge

Rising Fraud Rates

Sophisticated fraud attacks were increasing in frequency and complexity, with traditional rule-based systems unable to detect evolving fraud patterns.

Reactive Approach

Fraud was only investigated after occurrence, leading to customer fund losses and damaged trust in the bank's security measures.

Analyst Overload

High false positive rates created alert fatigue, with analysts overwhelmed by investigating legitimate customer activities flagged as suspicious.

Limited Visibility

Fragmented data sources made it difficult for analysts to get a complete picture of customer behavior and transaction patterns.

How might we proactively prevent fraud while maintaining a frictionless experience for legitimate customers?

Solution

I designed a comprehensive fraud analytics platform that combines AI-powered behavioral analysis with human expertise. The system learns normal customer patterns and flags anomalies in real-time, enabling analysts to prevent fraud before it impacts customers.

Solution Overview: AI-Powered Detection Interface

Design Process

1. Research & Discovery

I conducted interviews with fraud analysts, reviewed hundreds of fraud cases, and analyzed the existing detection workflow to understand pain points and opportunities.

User Research: Analyst Interviews & Workflow Mapping

Key Insights

  • Analysts spent 70% of their time investigating false positives
  • Context switching between multiple systems reduced efficiency by 40%
  • Lack of behavioral patterns made it difficult to distinguish fraud from normal activity

2. Ideation & Wireframing

Working closely with fraud analysts, I created wireframes that prioritized actionable insights and reduced cognitive load through progressive disclosure of information.

Early Wireframes: Dashboard Layout

Alert Detail View Wireframes

3. Testing & Iteration

I conducted usability testing with 8 fraud analysts, iterating on the design based on their feedback. The final design reduced average investigation time from 15 minutes to 4 minutes.

Key Features

Real-Time Behavioral Analysis

The AI engine monitors every transaction against learned behavioral patterns, instantly flagging anomalies while explaining the reasoning behind each alert.

Feature: Real-Time Detection Dashboard

Intelligent Alert Prioritization

Machine learning ranks alerts by risk score and confidence level, ensuring analysts focus on the most critical threats first.

Feature: Prioritized Alert Queue

Unified Investigation Workspace

All customer data, transaction history, and behavioral insights consolidated in a single view, eliminating context switching and reducing investigation time.

Feature: Investigation Workspace

Impact

The fraud analytics platform transformed Sterling Bank's approach to financial crime prevention, delivering measurable improvements across multiple dimensions while enhancing both analyst efficiency and customer trust.

78%
Fraud Reduction

Successful fraud attempts decreased dramatically through proactive detection and prevention.

€2.3M
Annual Savings

Combined savings from prevented fraud, reduced investigation costs, and lower legal fees.

92%
Analyst Satisfaction

Analysts reported significantly improved workflow efficiency and job satisfaction.

-73%
False Positives

Machine learning dramatically reduced alert fatigue by filtering out legitimate activities.

This platform has completely changed how we work. We're now preventing fraud before it happens, not just reacting to it. The time savings alone have been incredible.

Senior Fraud Analyst, Sterling Bank

Key Learnings

Design for Expertise, Not Simplification

Fraud analysts are domain experts who need powerful tools, not oversimplified interfaces. The key is organizing complexity through clear information hierarchy and progressive disclosure, not hiding it.

AI Transparency Builds Trust

Showing analysts why the AI flagged something as suspicious (behavioral patterns, risk factors) was crucial for adoption. Black box AI decisions were met with skepticism and resistance.

Context is Everything

Reducing context switching by consolidating information had a bigger impact on efficiency than any individual feature. Analysts needed to see the full picture in one place.

Iterate with Real Cases

Testing with actual fraud cases, not hypothetical scenarios, revealed critical edge cases and workflow needs that wouldn't have been discovered otherwise.