Reinventing Corporate Loan Underwriting: Insights from Annual Reports

This article is co-authored by Abhishek Sharma(Data Scientist), Misha Rana (Data Scientist), and Sampurn Rattan(Data Scientist) working at the Lumiq Research Team.

TL;DR: The Underwriting process for big corporate entities is a tedious task. The documents are huge, containing a combination of tabular and unstructured data which makes it rather time-consuming and laborious. Using Drishti Document AI, we have tried to transform the tedious process into a digitized, automated one transforming the way corporate underwriting is approached today.

Not too long ago, the Reserve Bank of India warned about a credit crisis brewing in July 2021. What’s alarming is that the RBI predicted that the Gross Non Performing Assets (NPA) ratio for Indian commercial banks would climb to 9.8–11.22% in March 2022, up from 7.48% in March 2021. SOURCE

To contain this problem, the Indian Banking Industry has made significant changes to its underwriting process. These changes include incorporating AI/ML technology to enhance their financial analysis strategies.

Let us look at a scenario in which a corporate bank is required to automate data extraction and processing from financial documents. The data extraction includes huge, complicated, and unstructured financial files like annual reports (ARs), stock statements, and stock audit reports. The underwriting team can then use the extracted data to assess a company’s financial performance comprehensively.

In this blog, we will focus on the corporate underwriting process needed to assess the health and Net Present Value of a company in order to make an informed credit decision.

Corporate banking is a subcategory of banking that includes a variety of services that are only available to businesses. The supply of credit is the most significant of these services. For this provision, all organizations must go through a lengthy and laborious corporate underwriting process. On the other hand, underwriting in retail banking is comparatively an easy process as it does not require huge amounts of corporate credit. Also, underwriters are being challenged to move from hindsight, where underwriting decisions are evaluated after the fact, to foresight, where portfolios are actively monitored, to understand the impacts of risks added to their books of business in a transparent, explainable, and real-time process.

An underwriter uses financial statements produced at regular intervals to assess a company’s financial health, risk profile, and market position. The annual report is the most important of such documents as it provides a bird’s-eye perspective of the company’s financials for the previous year.

When we surveyed a pool of underwriters, we could clearly see the uphill task laid in front of us. Pretty early in our research, we could see 2 primary challenges we were going to face in our pursuit.

1. No two annual reports are alike:

All annual reports are different as they are all tailored to the needs of the company, keeping in mind the industry in which they operate. Manually going through the complete document to find the information they need is akin to looking for a needle in a haystack.

2. All underwriting is subjective (as all risk is relative):

An underwriter’s job requires the capacity to compare multiple (financial) profiles. The underwriter, on the other hand, has no method of comparing two ARs from two different firms because each AR displays the same data in different formats, at different locations within the document, and with different meanings.

Why can’t we digitize the whole process?

Good question. The thing is that the traditional process of corporate underwriting is executed by a team of financial experts who use their brute force to go through the documents one at a time and manually choose the necessary metrics that can be used to make a credit decision. The obvious advantage is that they can identify the interchangeably used jargon that a person (or system) without the domain knowledge would not be able to identify. We were against a team of extremely smart and intelligent people. We realized this was our AlphaGo moment.

We came up with Drishti, Lumiq’s IDP Platform as a one-stop solution. Drishti Document AI, Lumiq’s AI-enabled intelligent document processing solution assists businesses to automatically extract huge amounts of data and process this information with ease. Drishti comes with built-in support for a variety of document types, which includes government ID cards, financial documents, and medical records. You can explore more about our solutions in the Agency Workplace, Customer Advisory, and Pryzm sections.

In the screen recording below, we can see how Drishti processes an annual report, page by page, and systematically analyzes all important information in this video:Video Link

When we implemented the project for a leading corporate bank, they could clearly see the benefits of automating the underwriting process compared to the traditional approach. We made the following observations

• 90% of line-of-credit was processed automatically without any manual assistance. This resulted in a 60% labor cost reduction and reduced the TAT by 100%.
• Drishti summarized a document that would normally be hundreds of pages into a single page. By providing such information in a single view, we successfully reduced the underwriting turnaround time and accuracy even while performing manual underwriting.
• Drishti also conducted a sentiment analysis of the bank’s financial documents to provide an overview for the company while also allowing the flexibility to compare corporations.

We have created a scalable, expandable, and incredibly flexible solution. This has expanded the far-reaching potential of converting unorganized financial data automatically.

Lumiq also has a simple user interface that allows individuals to be in the processing loop. Users can examine and modify data, get confidence scores of information extraction, and engage with the IDP in any way that they like. For more specialized solutions, you might want to look into Smart Credit and Insurance Underwriter tools available on the Microsoft Azure Marketplace.

To know more about Lumiq’s intelligent document processing capabilities, visit us athttps://lumiq.ai/drishti/

We also used Drishti to automate customer onboarding for one of the biggest insurers in India, to know how we achieved it, also read Automating customer onboarding with Lumiq Drishti.