Accelerating Mortgage Document Processing with Amazon Textract and Bedrock

Rocket Close, a Detroit-based title and appraisal management company within the Rocket Companies environment, announced on April 2, 2026, the deployment of a generative artificial intelligence solution to automate the processing of mortgage documents. The system, developed in collaboration with the AWS Generative AI Innovation Center (GenAIIC), reduces the time required to process abstract document packages from up to 10 hours per package to less than two minutes.

The company processes approximately 2,000 abstract document packages every day. Each of these packages averages 75 pages and contains complex legal and financial records related to property ownership and lending.

Technical Implementation and Performance

The automated workflow utilizes Amazon Textract for optical character recognition (OCR) and Amazon Bedrock for document analysis. Textract converts scanned documents into machine-readable text, while Bedrock employs foundation models to classify documents and extract specific data fields.

Amazon Bedrock is a managed service that provides a serverless method for building and scaling generative AI applications, offering a single API to access various foundation models. Rocket Close reported that this implementation has made the process 15 times faster.

The system achieves approximately 90% overall accuracy in field extraction, document classification, and segmentation. We see designed to scale to more than 500,000 documents annually.

Scope of Document Processing

The AI system processes more than 60 different document types. These include:

• Deeds
• Mortgages
• Liens
• Tax filings
• Court records

The workflow extracts structured data across categories such as legal judgments, ownership history, and loan details. This automation addresses operational challenges caused by inconsistent formatting, varying document structures, and the presence of handwritten notes in abstract packages.

Human Oversight and Verification

Despite the automation of data extraction, Rocket Close maintains human intervention to ensure the accuracy of the final records.

The human will step into almost all of the transactions, whether it’s verifying the data or looking at the exceptions, so that we’re ensuring that we’re doing the right thing for our clients and making sure that they have the proper homeownership rights to the property

Nathan Schrauben, chief information officer for Rocket Close

Nathan Schrauben stated that the company benchmarked Amazon Textract against other industry leaders and found that it seem to come out on top.