Finance ops · Production

Onity Group achieves 50% cost reduction and 20% accuracy improvement in mortgage document processing with Amazon Bedrock

The problem

Onity processes millions of mortgage servicing pages annually across hundreds of document types, but traditional OCR and AI/ML solutions could not reliably handle verbose legal text, inconsistent handwritten entries, notarization and legal seal detection, or limited contextual understanding.

First attempt

Traditional OCR and ML models proved fundamentally limited for mortgage servicing documents, failing on four core challenges: verbose legal text with buried data elements, inconsistent handwriting style variations, notarization and seal detection requiring visual understanding, and lack of semantic context interpretation.

Workflow diagram · grounded in source

Document upload to S3

trigger

“Documents are uploaded to Amazon Simple Storage Service (Amazon S3). Uploading triggers automated processing workflows.”

Document preprocessing

ai_action

“Before analysis, documents undergo optimization through image enhancement, noise reduction, and layout analysis. These preprocessing steps help facilitate maximum accuracy for subsequent OCR processing.”

Classification with confidence routing

routing

“Extracted content is processed by Onity's custom AI model. If the model's confidence score meets the predetermined threshold, classification is complete. If the document isn't recognized because the model isn't trained with that document…”

Dynamic extraction via Textract or Bedrock

ai_action

“It then dynamically routes extraction tasks between Amazon Textract and Amazon Bedrock FMs based on the complexity of the content. For example, verifying notarization requires complex visual and textual analysis. In these cases, the appl…”

Extracted data persistence

output

“The extracted information is stored in a structured format in Onity's operational databases and in a semi-structured format in Amazon S3 for further downstream processing.”

Reported outcome

Onity achieved a 50% reduction in document extraction costs and a 20% accuracy improvement compared to their previous OCR and AI/ML solution.
Home appraisal checklist review accuracy improved by 65% over the manual process, and credit report analysis achieved approximately 85% accuracy.

Reported metrics

Document extraction cost reduction50%

Overall accuracy improvement vs previous solution20%

Home appraisal review accuracy improvement vs manual65%

Credit report analysis accuracyapproximately 85%

Reported stack

Amazon BedrockAmazon TextractAnthropic's Claude SonnetAWS KMS

Source

https://aws.amazon.com/blogs/machine-learning/automating-complex-document-processing-how-onity-group-built-an-intelligent-solution-using-amazon-bedrock?tag=soumet-20

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Onity achieved a 50% reduction in document extraction costs and a 20% accuracy improvement compared to their previous OCR and AI/ML solution.

What tools did this team use?

Amazon Bedrock, Amazon Textract, Anthropic's Claude Sonnet, AWS KMS.

What results were reported?

Document extraction cost reduction: 50%; Overall accuracy improvement vs previous solution: 20%; Home appraisal review accuracy improvement vs manual: 65%; Credit report analysis accuracy: approximately 85% (source-reported, not independently verified).

What failed first in this deployment?

How is this finance ops AI workflow structured?

Document upload to S3 → Document preprocessing → Classification with confidence routing → Dynamic extraction via Textract or Bedrock → Extracted data persistence.

Onity achieved a 50% reduction in document extraction costs and a 20% accuracy improvement compared to their previous OCR and AI/ML solution.Home appraisal checklist review accuracy improved by 65% over the manual process, and credit report analysis achieved approximately 85% accuracy.