Transforming Document Processing in Mortgage Servicing: Onity Group’s Journey with Amazon Bedrock and AWS

Introduction to Onity Group and the Significance of Efficient Document Processing

The Challenges of Traditional OCR and AI Models in Mortgage Servicing

Innovative Solutions: Onity’s Intelligent Document Processing Approach

Ensuring Security in Document Handling for Financial Services

Case Studies: Enhancing Efficiency with Amazon Bedrock

Conclusion: A New Era in Document Processing for Onity Group

About the Author

Transforming Document Processing: Onity Group’s Journey with Amazon Bedrock

In the dynamic landscape of the mortgage servicing industry, efficient document processing is not just a matter of convenience; it’s crucial for business growth. In this post, we delve into how Onity Group, a financial services company specializing in mortgage servicing and origination, has revolutionized its document processing capabilities using Amazon Bedrock alongside other AWS services.

Onity Group: A Brief Overview

Founded in 1988 and headquartered in West Palm Beach, Florida, Onity Group operates through its primary subsidiary, PHH Mortgage Corporation, and the Liberty Reverse Mortgage brand. The company provides comprehensive mortgage servicing and origination solutions to homeowners, business clients, and investors, handling millions of pages across hundreds of document types annually.

However, Onity faced notable challenges: processing legal documents like deeds of trust, which often buried critical information within dense text, inconsistent handwritten entries, and verification of notarization and legal seals—tasks where traditional optical character recognition (OCR) and AI/ML solutions struggled.

A Game-Changer: Amazon Bedrock

To meet these challenges, Onity adopted foundation models (FMs) offered by Amazon Bedrock, achieving a 50% reduction in document extraction costs and bolstering overall accuracy by 20%. Onity’s intelligent document processing (IDP) solution dynamically routes extraction tasks, effectively utilizing its custom AI models alongside AWS’s generative AI capabilities.

Raghavendra (Raghu) Chinhalli, VP of Digital Transformation at Onity Group, emphasizes the need for a solution that evolves rapidly alongside document processing demands. Priyatham Minnamareddy, Director of Digital Transformation & Intelligent Automation, added, “By combining AWS AI/ML and generative AI services, we achieved the perfect balance of cost, performance, accuracy, and speed to market.”

The Limitations of Traditional Document Processing

Several fundamental challenges prompted Onity’s search for advanced solutions:

• Verbose Documents: Key mortgage servicing documents contain dense text with critical information obscured within complex structures, such as identifying legal descriptions in a deed of trust.

• Inconsistent Handwriting: Handwritten elements vary significantly, throwing a wrench into the extraction process. Simple variations (e.g., state names) can create substantial challenges.

• Notarization and Legal Seal Detection: Identifying notarized documents and detecting legal seals often requires nuanced understanding beyond traditional OCR capabilities.

• Limited Contextual Understanding: Traditional models excel at digitizing text but struggle to interpret the semantic context, hindering true information extraction.

These challenges drove Onity to seek sophisticated solutions that traditional OCR and ML models simply couldn’t address.

The Intelligent Solution

Onity’s refined approach harnesses AWS AI/ML and generative AI services. Core components of the solution include:

• Amazon Textract: This ML service automates text extraction from documents, streamlining workflows and unlocking valuable data.

• Amazon Bedrock: This fully managed service grants access to high-performing FMs, enabling Onity to flexibly choose models that balance cost, accuracy, and performance.

How the Solution Works

The document processing workflow includes several crucial steps:

1. Document Ingestion: Documents are uploaded to Amazon S3, triggering automated processing.

2. Preprocessing: Documents undergo enhancement, noise reduction, and layout analysis to optimize OCR accuracy.

3. Classification: A three-step intelligent workflow identifies document types using Amazon Textract and Onity’s custom AI model, routing unrecognized documents to Amazon Bedrock’s models for further classification.

4. Extraction: The solution dynamically routes extraction tasks to Amazon Textract or Bedrock models based on content complexity. For instance, verifying notarization requires advanced analysis, thus tapping into the capabilities of Amazon Bedrock.

5. Persistence: Extracted information gets structured storage in Onity’s operational databases and semi-structured formats in Amazon S3 for downstream processing.

Security Measures

Given the sensitivity of financial documents, Onity employs stringent data protection measures. Data is encrypted at rest and in transit, and access is tightly controlled. Compliance with AWS security best practices ensures that their architecture remains robust and secure.

Transformative Use Cases with Amazon Bedrock

Onity’s implementation showcases how Amazon Bedrock can revolutionize document processing in various ways:

1. Deed of Trust Data Extraction

Deeds of trust often contain vital information within verbose legal text. Onity’s solution reduces extraction costs by 50% while enhancing accuracy by 20% compared to previous methods.

2. Checklist Review for Home Appraisals

An automated solution reviews property appraisals for discrepancies in data points, improving review efficiency by 65% and reducing errors.

3. Automated Credit Report Analysis

Credit reports, with their disparate formats, previously posed extraction challenges. The solution standardizes credit scores across formats, achieving an accuracy of approximately 85%.

Conclusion

Onity Group’s intelligent document processing, bolstered by AWS generative AI services, demonstrates how organizations can turn complex document handling obstacles into strategic advantages. With a remarkable reduction in costs and significant accuracy improvements, Onity’s innovative approach provides a roadmap for other businesses seeking to modernize their document processing operations.

For organizations grappling with similar challenges, starting with Amazon Textract can unlock its core capabilities. For more sophisticated use cases, exploring Amazon Bedrock model options may yield substantial benefits.

As Ramesh Eega, a Global Accounts Solutions Architect, notes, leveraging modern AI technologies can fuel innovation and drive efficiency in document processing, ensuring businesses remain competitive in an ever-evolving market.