Advertisement

  • News
  • Columns
  • Interviews
  • BW Communities
  • Events
  • BW TV
  • Subscribe to Print
  • Editorial Calendar 19-20
BW Businessworld

Driving Automation And Efficiency In Loan Processing Using Text Analytics

The text analytics tool can process documents like PAN Card, Bank Statements, Passport, Voter ID, Aadhar, CIBIL score report etc. and extract information relevant to credit worthiness assessment in real-time.

Photo Credit :

1598595018_eL2qKz_KSE_IMAGE_2.jpg

Text analytics can help businesses across domains; listen to the right stories by extracting insights and identifying patterns and trends. 

Let’s take a closer look at how a text analytics product improves the entire loan processing exercise. From the time a customer fills up the loan application to the time it is approved and disbursed, the entire process can take anywhere up to 7-8 days. There are several formalities to be observed, right  from collecting loan related documents such as KYC, financial statements, tax returns and so on.

Across the lending lifecycle, here are the business processes which can be automated using a text analytics tool.

Auto-KYC & Creditworthiness assessment in a jiffy

One of the most important steps, which is also the most time consuming, understands the creditworthiness of the customer, assessed from the documents such as KYC, financial statements and tax returns. This process is also important as it provides insights into the ability of the borrower to pay back a loan which is the biggest factor in determining whether a loan can be granted or not.

The text analytics tool can process documents like PAN Card, Bank Statements, Passport, Voter ID, Aadhar, CIBIL score report etc. and extract information relevant to credit worthiness assessment in real-time. This extracted information then goes in a modelling pipeline to create a baseline model. The baseline model is augmented by credit history information to decide on a final result whether to give a loan or not. All this happens within minutes.

Making robust credit decisions using unconventional Sources i.e. Alternative data

Lending institutions can leverage predictive analytics and natural language processing to expand the scope of verifying the creditworthiness of borrowers and reduce loan defaults.

Instead of tracking just a few numeric variables to determine credit scores, machine learning models can tap into unconventional sources such as social media posts, geolocations, browsing activities, utility bill payments etc. to generate credit scores using a larger number of variables. Text data extraction from images, videos, and text files can provide deeper insights to help establish patterns and categorize borrowers as those who are likely to pay versus those who tend to default.

According to research by Bloomberg, digital lending platforms will touch a market size of US $15.3 billion by 2026. One of the key drivers of this digital lending revolution will revolve around lending analytics. The key to lending analytics revolves around the ability to embrace text analytics and automate key processes around extracting data from unstructured text-based inputs and then converting this data to actionable insights.

A bulwark against Fraud Detection

The second area where AI-based systems can help is in fraud detection by borrowers who stack multiple loans. The AI system can detect when an individual or organization applies for multiple loans from different lenders. This can be spotted by analyzing key financial documents like bank statements, balance sheets (in the case of an enterprise), and such.

The lenders can also drive timesaving on due diligence efforts by deploying an AI-based system that can scan several sources to gather and update data and prevent frauds.

Of course, a risk team with human intelligence is absolutely essential, especially in case of larger loans, but fraud detection using AI can eliminate many potential risky profiles at the first level of scrutiny.

Text analytics tools help here by rendering the bank statements and balance sheets and other documents analytics-ready by converting them to CSVs or JSONs from PDFs or images. Improving Cross Sell and its Due Diligence

Once the loan has been disbursed, the technology can be used effectively to assess the behavior of the customer and either upsell or caution depending on the situation using recommendation system methods. This can help make the underwriting process more efficient for the cross-sell process. The information-backed due diligence helps to make better decisions around managing risk and alerting collection teams to ensure timely loan repayments.

Document Digitization of paper-based information

Paper-based forms are still filled and scanned and stored as images for a lot of processes in retail banking. To browse through them and get the relevant data can be difficult if done manually. Document digitization of such paper-based documents and tagging them with metadata can speed up the lending process. With AI tools, this can be automated and data extraction made easier. This is especially useful for accessing and tracing documents, critical for compliance and audits.

The future of Text Analytics

The Text Analytics Market size soared past USD 5 billion in 2019. It is expected to grow at 18% CAGR over the period of 2020-26 meaning it will grow to over USD 15 billion by 2026. Owing to this, it is not surprising that, as companies look to put their best foot forward, they are increasingly turning to text analysis to improve customer experiences and business processes. For example, if a Fintech firm uses an AI powered Text Analytics solution to automate the process of assessing a customer’s creditworthiness, the solution can also enhance processing time and avoid bad loans & NPAs by over 40% compared to the previous year where text analytics was not used.

Text analytics is developing a solid foothold, and it will continue to be a necessity in the coming years.

Disclaimer: The views expressed in the article above are those of the authors' and do not necessarily represent or reflect the views of this publishing house. Unless otherwise noted, the author is writing in his/her personal capacity. They are not intended and should not be thought to represent official ideas, attitudes, or policies of any agency or institution.


Mr. Ashish Kumar

Principal Data Scientist, teX.ai, Indium Software

More From The Author >>