Executive Summary
DealNet Capital engaged DataArt to design and implement a customer-facing web portal. The new solution leveraged state-of-the-art AWS services to automate and streamline a complex and time-consuming manual document review process. As a result, DealNet significantly improved its operational efficiency and customer experience through the loan application journey.
Why AWS
DealNet had previously utilized EC2 instances on AWS for its Dealer Portal, where homeowners and contractors submitted the required documentation for the audit process. Leveraging the same platform and expanding DealNet's AWS infrastructure was the logical choice. The broad range of available AWS services like Amazon Textract, Amazon Rekognition, Amazon SageMaker, etc., perfectly addressed the team’s main challenge: automating multi-step workflows with numerous repetitive manual steps.
DealNet and DataArt: Strategic Growth Partnership
DealNet’s prior relationship with DataArt as a strategic software engineering partner enjoyed a strong track record of success, mainly contributing DataArt’s broad technology expertise, previous experience with Business Process Management systems, custom automation, and DataArt’s status as an AWS Advanced Consulting Partner.
Meeting the Challenge
The DealNet-DataArt team, in collaboration, designed and implemented the new web platform leveraging the AWS cloud (Amazon A2I, Amazon Textract, and Amazon Rekognition). DealNet’s new Dealer Portal captures all new credit applications from consumers and dealers.
The re-implemented credit approval process supported by the platform is now extremely quick and consists of 3 simple steps:
- Scanning the customer’s ID populates the loan or lease application.
- In a few seconds, the system’s back-end provides the credit decision.
- Once the application is approved, the customer submits the package of mandatory documents that require validation for compliance, lending, and financial standards to get funded.
Partner Solution
Upon document submission for validation, the funding documents are processed by Amazon Textract to extract and identify key-value pairs and OCR text accurately. The extracted values are then compared against the validation dataset to ensure compliance.
Using the extracted key-value pairs and custom logic, the system identifies the location of customers’ signatures on loan and lease agreements. These signatures are then sent to Amazon Rekognition, which returns a confidence score to ensure the document is signed properly.
In cases where the extracted data does not match the expected results from the back-end system, the Amazon A2I initiates the human loop and sends data review tasks to the human staff.
The solution also provides comprehensive reporting on the document validation status, the number of documents in the human review queue, the percentage of papers reviewed with Amazon Textract, and other relevant metrics. These statistics enable management to make data-driven decisions regarding scaling the human workforce and identify bottlenecks and areas for potential workflow improvements.
To fulfill key business requirements, the portal design leverages the following AWS services:
- Amazon Textract:A fully managed machine learning service that accurately reads, processes, and extracts printed text, handwriting, forms, tables, and other data from any document without manual effort or custom code.
- Amazon Rekognition audits the documentation whether it contains signatures.
- Amazon Rekognition: Audits the presence of signatures in the documentation.
- Amazon SageMaker (Augmented AI): Used for manual document audits. If auto audit fails, Amazon A2I creates a human loop for manual document processing.
- Amazon S3: Stores contracts, files for AWS Textract and AWS Rekognition, and results from AWS SageMaker.
- Amazon SQS: Enables communication between the web portal and the document audit workflow.
- Amazon EC2: Used to run the web portal and the document audit service.
