INNOVYA TECHNOLOGIES

Case Study 2

For a significant US bank, TRUBOT (RPA) AND TRUAI (ARTIFICIAL INTELLIGENCE) FAST-FASTER DOCUMENT SEARCH AND RETRIEVAL OF 35+ MILLION PAGES.

Services

To combine the mortgage procedures of six acquired banks and to process, store, and retrieve mortgage documents through a single DMS powered by FileNet.

Industry

BSFI

The Challenge

Huge load of mortgage papers:

To digitally transform more than 1.8 million unstructured mortgage documents totaling 35+ million pages, while enhancing their searchability and retrievability by automatic analysis, indexing, and classification into 275 categories.

Unstructured documents:

To automatically classify documents that would take months to manually classify and are available in a variety of formats, including paper, scanned, and prescanned versions with medium to low resolution and stored on shared drives.

THE SOLUTION

In order to expedite and process the mortgage paperwork for the six acquired banks as a single entity, the client required 42 days to digitise, condense, and categorise 35+ million pages into 275 preset categories. After thoroughly examining their IT infrastructure, Innovya proposed Intelligent Data Capture, Robotic Process Automation (RPA), and Artificial Intelligence (AI) solutions, each of which was powered by their own proprietary products, TruCap+, TruBot, and TruAI. The following was the solution: 

 

Digitization: 
Utilising TruCap+’s Intelligent Data Capture engine, massive amounts of free text, unstructured documents may be turned into digital assets. 

Auto-analysis of document and metadata: 
Using TruCap+ to conduct a context-sensitive analysis of the documents and extract crucial fields, like loan number, loan date, amount, customer name, unique identification number, address, etc. 

Indexation: 
To automatically index documents using the key fields that were extracted 

Classification: 
To use an AI engine called TruAI combined with specialised evolutionary algorithms powered by NLP to automatically categories and classify the digital assets into 275 pre-defined categories OR to place them in a suspect folder for manual classification. 

Workflow: 
to assemble every component and group them into “classified,” “unclassified,” and “suspect” buckets. 

Exception handling & Machine Learning (ML): 
To classify the remaining suspects and unclassified documents, and to retrain the AI/ML algorithms using subject-matter experts so that they get better over time. 

Export: 
To automatically transmit the automatically classified documents, together with a file log, to the DMS powered by FileNet 

Conclusion

  • 35 million papers were automatically indexed and categorized into 275 categories. 
  • Within 42 days, a paper-based workflow was transformed into a digital framework, speeding up the business process and automatically classifying the mortgage papers. 
  • Shortened the time it takes to seek for and get documentation from 48 hours for each mortgage case to a few minutes. 
  • At least 150 man hours were saved each month in each location. 
  • Increased accuracy by 87%; the classification was substantially better than that performed manually by operators. 
  • 50% less operational expense was spent on physical storage, searching, rearranging, etc.