How Can You Accelerate Bulk Document Processing, Leveraging an AI Based OCR Tool?

By December 13, 2018 November 25th, 2019

From a borrower standpoint, they view loan processing as a lot of paperwork. It definitely is true as the documentation requirements related to financial transactions like mortgages are quite extensive and have a certain level of complexity associated to it. As a lender, you need to manage all this paperwork and if you want to get this done, fast and in an accurate manner, there is no other way than to rely on technology. Optical character recognition (OCR) technology can help in many ways, but it has had its own limitations in the past.

Originating bulk loans

In the case of wholesale lenders, in most instances, they need to originate bulk loans that they receive from brokers. Using the right technology, these bulk loan portfolios can be typically indexed and stacked, so that they can be fed into the Loan Origination system (LOS). This task is too difficult to manage manually. Manual entries can mean millions of pages that must be reviewed, indexed and analyzed. For instance, processing 2000 loans (each loan file with up to 500 pages), would add up to 1,000,000 pages for review and indexing! There is no way manual indexing can finish this job. And even if handled it would be a labour intensive task highly prone to delays and errors.

This is where the usual Optical Character Recognition (OCR) or Intelligent Character Recognition (ICR) technologies come into the picture. As automated as it is, there still is the necessity to create templates and rules outlining the data extraction patterns for each different document design processed. Very often this doesn’t pan out as envisaged and there is often a margin of error that can bear severe consequences.

There are other challenges too that have dampened OCR adoption in general.

  • The constant change of regulations, forms, and procedures- which means new templates and document design too.
  • Very often lenders lack the expertise in handling OCR software.
  • Extracting data using OCRs is usually an 8 to 10 step process and even the best of the OCRs do not have a data extraction efficiency of more than 70 to 80%.

Bulk document processing requires accuracy and where OCR fails, the inclusion of AI and NLP based tools for the same can enable accurate output. This can significantly help in avoiding rework and reducing the related risks. In addition, the work also gets completed faster.

VisiLoanReview An AI Based Tool for Mortgage

Visionet, brings to you VisiLoanReview, an Artificial Intelligence (AI) based classification and OCR tool, which has the capability to learn. It leverages the organization’s 10+ years of operational experience to configure the OCR engine for specific mortgage business rules, which result in accelerating the performance.


VisiLoanReview uses triple pass methodology to strip and validate content using NLP, OCR and multiple big data stacks to provide 99.5% guaranteed accuracy at lightening speeds. This AI-based system trains itself using Machine Learning recognition pattern algorithms from documents and multi-layered OCR formulations. If the image quality is extremely poor Visionet provides manual help to cleanse the data and documents as well.

Visionet has been able to provide high impact benefits to 50+ lenders. Industry first ZERO COST technology, along with our world-class service infrastructure makes it possible. For more information, feel free to reach out to us.

Sameer Wadhwa

Sameer Wadhwa

Sameer is a market facing client relationship leader with 18+ years of experience of working with clients. He has been focused on implementing Digital Transformation solutions for client’s businesses and he has specific focus on Mortgage and CPG domains.

Leave a Reply