Rocket Close transforms mortgage document processing with Amazon Bedrock and Amazon Textract
Using Amazon Bedrock and Textract, the system processes 2,000 complex document packages daily with 90% accuracy.
Rocket Close, a title and appraisal management company within Rocket Companies, has automated one of the mortgage industry's most tedious bottlenecks. The company was manually processing approximately 2,000 abstract document packages daily, with each file averaging 75 pages of complex legal documents. This manual extraction took an average of 10 hours per package, creating a massive operational drag and costing millions annually. In partnership with the AWS Generative AI Innovation Center, Rocket Close developed a solution using Amazon Textract for optical character recognition (OCR) and Amazon Bedrock to access various foundation models (FMs) for intelligent analysis.
The new system transforms this workflow, making it 15 times faster. It now takes roughly 40 minutes to process a package that previously required 10 hours of manual labor. The AI-powered pipeline handles over 60 different document classes—from mortgage agreements to chain of title deeds—achieving a strong 90% overall accuracy rate in document segmentation, classification, and critical field extraction. Designed to scale to over 500,000 documents annually, this automation eliminates human error inconsistencies and frees up significant resources. The implementation demonstrates how generative AI can tackle document-intensive processes in regulated industries, enabling faster customer service and sustainable growth by turning a cost center into a strategic advantage.
- Processes 2,000 complex abstract packages daily (each ~75 pages), scaling to 500k+ documents annually.
- Reduces manual extraction time from 10 hours to ~40 minutes per package—a 15x speed increase.
- Achieves 90% accuracy in segmenting, classifying, and extracting data from over 60 document types using Amazon Bedrock FMs.
Why It Matters
Shows how generative AI can automate high-volume, complex document workflows in regulated industries, turning major cost centers into efficiency engines.