HOW DOES IT WORK

Document upload process starts in a web application.

Uploaded documents such as insurance claims or mortgage applications in various forms such as emails.

The document extraction and classification workflow utilizes: Amazon S3 Amazon SQS, AWS Lambda, Amazon Textract, Amazon SNS, Amazon Comprehend Custom Classifier, Augmented AI – A2I (Human in loop), Amazon Dynamo DB.

Workflow uses a combination of OCR & NLP to extract text from documents and understand the insights they contain.

HOW CAN YOU GET STARTED

RAPID ASSESSMENT

• Understand needs, objectives, and existing document workflow
• Collect and review documents, perform exploratory data analysis
• Review current infrastructure and TCO
• Ideate and prioritize potential solution
• Select representative subset document types for prototyping
• Define success factors including accuracy and latency required
• Advise on next steps

RAPID AI PROTOTYPING

• Scope out prototyping
• Design proposed new architecture
• Set up document processing environment
• Annotate documents
• Prototype document processing pipeline
• Evaluate document classifying and extraction accuracy
• Evaluate human-in-the-loop needs
• Determine dependencies and need for upskilling
• Confirm success factors
• Calculate total cost of ownership
• Define path to production
• Final report