Increasing Efficiency While Maintaining Effectiveness in Credit Underwriting | A Tale of Incorporating AI into Traditional Financial Processes

Despite the vast amount of change seen in financial services in recent years, credit underwriting and risk management processes remain largely unchanged, especially at legacy institutions. With innovation constrained in part by compliance regulations, many banks simply choose to leave things alone unless the regulations guiding the process change.

Fintechs can better incorporate technology into underwriting processes while still adhering to all relevant regulations. Innovation is part of the overall ‘fintech’ brand and companies building a process from the ground up don’t have to deal with “this is how we’ve always done it here” roadblocks.

Most recently, fintechs have begun to incorporate AI into their underwriting processes. Torpago Director of Risk, Brendan Coons, shared how AI  builds on the solid foundations of legacy institution processes to make underwriting and risk management faster and more efficient.

Traditional Underwriting and Risk Mitigation Tactics

In addition to evaluating the financials of a potential borrower, lenders also have to ensure the applicant is not subject to any regulatory restrictions. At the same time, they have to make sure that their process is fair and equal. “In the commercial space, the big ones are KYC and AML, making sure that you're not lending to or servicing customers that are involved with terrorism or illegal activities,” said Coons when highlighting critical regulations that need to be followed. “And then there's the Equal Credit Opportunity Act and Regulation B, where we need to send adverse action notices to customers when needed.”

In traditional underwriting, dedicated teams often perform these tasks manually and take place early in the application processing stage. Speaking about his experience at a prior large commercial bank, Coons said “It was a very manual process. We had a whole team doing it and that was upfront – before we did any financial analysis." Steps in the process include:

  • Verifying active business entities and confirming organizational ownership.
  • Checking against sanctions and compliance lists for due diligence.
  • Financial analysis through manual review of provided financial statements, including business tax returns and audited statements.
  • Creating "spreads" for year-over-year financial performance analysis and risk score determination.

The entire process from application to decision-making could take at least a business week. Additionally, if an application is approved, ongoing monitoring has to occur, including scheduled annual reviews of customer credit and reassessments of financial information to validate or adjust credit limits.

How Torpago Leverages AI for Underwriting

When building our underwriting process, our primary objective was to increase efficiency while upholding the effectiveness of a bank underwriting process. 

Our innovative approach begins with automating the initial stages of underwriting, particularly in areas like KYC, AML, and business entity verification. This is accomplished by screening the business data provided in the application against government sanctions lists and through fraud databases. In cases where information cannot be verified or is high risk, the Torpago team gets in touch with the applicant and potentially schedules a video conference to verify their identity. 

Automating the verification process for the majority of applicants significantly shortens application processing time while retaining accurate anti-fraud and compliance checks. “If it can make it through compliance with no red flags, that process takes place in a matter of seconds. And then that allows us to feel confident that the financial data that we're getting from them can be relied on to run our credit risk models,” said Coons.

For credit approval purposes, we leverage sophisticated AI models to analyze financial data. “We collect thousands of bank transaction data points from the business’ bank account and then use machine learning to run analysis and determine creditworthiness,” Coons explained.

“We look how much cash flow they have coming into their account each month, are they covering their debt payments, is the company generating organic cash from operations or using debt versus equity to support operations? We're also looking at their average cash balance trends over time and how many non-sufficient funds (NSF) events they have had. We're gathering all of these data points for risk scoring and to assign dynamic credit limits."

This application of AI is extended into our ongoing monitoring and fraud detection processes, where we use it to detect unusual transaction patterns. As long as we have that live connection to the business’ bank account, we can access customer data in real time and have created an alert driven underwriting system. “We even have an Underwriting Bot and Fraud Bot which alert us on suspicious activity or declining risk; it's really pretty cool!” said Coons.

Torpago has started to explore how AI can help analyze unstructured data, such as news articles, economic reports and company websites, through natural language processing capabilities. This can help underwriters gauge the overall sentiment surrounding a business or industry, and provide them with company and management background.

Recommendations for Banks and Credit Unions

Coons had some suggestions for banks that are interested in modernizing their underwriting processes and adding AI to their tool chest. “It's really about building efficiency while maintaining or improving on their current level of effectiveness,” he suggested. “By giving them access to the data, they are better able to leverage AI in their own process.”

We are monitoring payment success rate, early stage delinquency and fraud rates to gauge the effectiveness of our underwriting. We have controls where accounts that fail payments are suspended, and if they develop a pattern for late payments, we factor that into the risk assessment process.

“I think of it as 80% automated and 20% manual and human-involved. If you think of credit limits, you want to dedicate your time and effort towards the 20%, higher limit accounts, that could really move the needle if one were to go bad."

Start Modernizing With Powered By

Our range of Powered By solutions were designed to help community banks and credit unions modernize their products and processes, making them more competitive and profitable. With our business credit card and spend management platform, financial institutions can launch a modern credit program, supported by our underwriting, marketing, and customer support teams, that attracts new customers and delights existing ones. Get in touch today to learn more!

Request a Demo

Back to Blog