AI-Powered Collections: How EXL and AWS are Transforming Debt Recovery

Financial institutions are facing increasing pressure to improve collections operations. Traditional methods are proving less effective, leading to higher costs, poor customer experiences, and increased regulatory scrutiny. EXL Service Holdings has partnered with Amazon Web Services (AWS) to develop PayMentor, an AI-powered platform designed to personalize customer outreach, optimize communication channels, and automate workflows at scale, addressing these challenges head-on.

The Challenge in Modern Collections

Legacy debt collection strategies often rely on a one-size-fits-all approach – aggressive phone calls, generic messaging, and rule-based workflows. However, customers have varying capacities to pay, willingness to pay, and channel preferences. This results in rising collection costs as phone-based strategies become less effective, a decline in customer satisfaction impacting Net Promoter Score (NPS), and potential regulatory issues from bodies like the Consumer Financial Protection Bureau (CFPB) and state regulators. Scaling contact centers to meet growing loan portfolios also presents a significant constraint.

Introducing PayMentor: An AI-Driven Solution

EXL recognized that effective collections require a data science and customer engagement approach – determining who to contact, when to contact them, how to contact them, and what to say. PayMentor, built on AWS, leverages machine learning to deliver personalized experiences and automate communication, supporting millions of daily customer interactions.

Why AWS for PayMentor?

EXL chose AWS for five key reasons:

• Financial Services Compliance: AWS provides robust compliance certifications (PCI-DSS, SOC 1/2/3, ISO 27001), a comprehensive suite of services, and support for regulatory audits.
• Production-Ready ML Infrastructure: Amazon SageMaker offers scalable model training and inference, isolated environments for development and production, and built-in security controls for financial data.
• Omnichannel Communication Services: Native integrations with Amazon Simple Email Service, AWS End User Messaging, Amazon Connect, and Amazon Lex, coupled with pay-per-utilize pricing and a global footprint, enable flexible and cost-effective communication.
• Real-Time Data Processing: Amazon Kinesis enables streaming customer engagement data, low-latency feedback loops for strategy optimization, and event-driven architecture with Step Functions.
• Scalability and Cost Efficiency: Serverless architecture using AWS Lambda eliminates over-provisioning and auto-scales to handle fluctuating interaction volumes.

PayMentor’s Architecture: A Phased Approach

PayMentor’s architecture is designed with security, compliance, resilience, scalability, and cost optimization in mind. The solution is implemented in phases:

Phase 1: Secure Data Ingestion

PayMentor supports three data ingestion methods:

• SFTP: For batch data, utilizing end-to-end encryption and AWS Glue ETL for data validation and loading into an Amazon S3 data lake.
• REST API: Leveraging AWS Lambda and Amazon API Gateway for serverless, auto-scaling capabilities, with authentication via Amazon Cognito and protection via AWS WAF.
• Third-Party Integrations: Using Amazon AppFlow for secure, managed integrations with credit bureaus and payment processors.

Phase 2: ML Model Development (Isolated Account)

To maintain security and compliance, EXL separates ML development from production using isolated AWS accounts. This ensures data security, auditability, risk management, and governance.

PayMentor utilizes two primary ML models:

• Customer Ranking Model: Predicts the probability of payment using Gradient Boosted Trees (XGBoost on Amazon SageMaker) and is retrained weekly.
• Channel Preference Model: Predicts the optimal communication channel based on historical engagement data from Kinesis streams.

Phase 3: Daily Scoring and Strategy Generation

AWS Step Functions orchestrates SageMaker inference to score accounts daily and generate personalized communication strategies. EventBridge triggers the workflow, which fetches data from S3, invokes SageMaker, saves rankings to RDS, generates communication plans, and sends notifications via Amazon Simple Notification Service.

Phase 4: Omnichannel Execution

AWS Lambda orchestrates channel-specific services for personalized communication:

• Email: Amazon SES for payment reminders and account summaries.
• SMS: AWS End User Messaging for urgent notifications and payment links.
• WhatsApp: Third-party services integrated with Lambda.
• Voice IVR: Amazon Connect with Amazon Polly and Amazon Lex for automated payment arrangements.
• Voice AI: Amazon Connect with Amazon Lex and Amazon Bedrock for natural language conversations.
• Self-Service Portal: API Gateway with Lambda for customer-initiated account management.

AWS Translate supports multi-language messaging.

Phase 5: Real-Time Engagement Tracking

Amazon Kinesis streams capture SMS, email, and voice interactions in real-time. Data Firehose sends data to S3 for archival and Kinesis Data Streams for processing, updating Amazon RDS and feeding Amazon QuickSight for visualization.

Phase 6: Analytics and Reporting

Amazon QuickSight dashboards, powered by Amazon RDS and Amazon S3, provide real-time visibility into key metrics, including message delivery rates, engagement rates, cost per contact, cure rates, and regulatory compliance.

Security and Compliance

PayMentor prioritizes data protection through encryption (AWS KMS), network security (VPC isolation, AWS WAF), and robust authentication and authorization (Amazon Cognito, AWS IAM). SageMaker security controls include endpoint encryption, restricted notebook access, and VPC isolation.

Conclusion

The future of collections lies in intelligent engagement. AWS provides the scalable, secure, and compliant infrastructure needed to deliver personalized experiences and drive measurable results. For financial institutions looking to modernize their collections processes, AWS offers a powerful and comprehensive solution.

To learn more about how AWS is transforming financial services, explore AWS for Banking.