Can Artificial Intelligence Jeopardize Banks’ Loan Growth?

More banks and credit unions are turning to fintech platforms to digitally transform their loan software. But, if a financial institution isn't running the right model to determine the profitability of its portfolio, they could be losing money and not know it.

Using artificial intelligence and automation to digitize and upgrade banking processes is the new imperative for banks and credit unions. They have come to realize that becoming truly digital institutions requires much more than offering a mobile banking app, or turning paper files into digital files. Digital transformation requires a much more comprehensive approach.

One of the most challenging areas is in lending, the bread and butter of banking. Where loan originations used to take place in a face-to-face meeting between a banker and client — taking upwards of two to three hours (depending on the type of loan) — they are now taking place online and on mobile phones.

That’s great in many ways for both financial institutions and customers. However, as banks and credit unions have begun to embrace lending platforms that incorporate artificial intelligence and loan origination algorithms, institutions may be setting themselves up for profitability downfalls if they aren’t using the right set of capabilities and metrics. The experience of Pennsylvania-based Huntingdon Valley Bank offers key insights.

“Previously, we were using manual-processes — like much of the banking industry — to understand loan pricing and profitability models, which is a completely outdated approach,” says Hugh Connelly, EVP and Chief Lending Officer at the bank.

Food for Thought:

Banks are automating front-end processes, but they should use technology to automate loan pricing.

Huntingdon Valley, with assets of $557.8 million, is an example of a community bank weaving technology into its backend systems. To help it transform its operations, the bank tapped fintech vendor nCino to automate its loan origination software. The tech company’s cloud-based platform incorporates artificial intelligence and machine learning for several functions including leveraging predictive analytics to measure performance and monitor risk across the organization.

To infuse artificial intelligence technology into a bank’s underwriting process requires knowing which profitability models work (and which don’t). Done properly it can be a competitive advantage for community and regional banks and credit unions.

Two Key Lending Functions Ripe for Automation

In the shift to digital processes, the two primary issues many banks and credit unions are trying to upgrade are loan pricing and loan profitability. Connelly points out that many banks still do this manually, which makes it difficult to grow loans at scale.

“It’s basically business 101: Make sure you know how to make money,” Connelly jokes. “Yet, banks have historically struggled with it.”

Every financial institution’s needs are different, says Will Cameron, Senior Vice President of Community and Regional Banking at nCino. Even in shifting from a legacy paper model to a digital one, the artificial intelligence technology that a bank builds or buys must be adaptable to their pre-existing structure and budget.

“The challenge for a vendor is to make a product that is configurable enough to meet everybody’s different needs,” he explains. “We mold ourselves around the particular methodologies that people have.”

Read More: Banks Toy With Dumping Credit Scores from Lending Decisions

Applying Artificial Intelligence to Loan Profitability

There are three primary methods banks use to measure the profitability of their products. A popular one is risk-adjusted return on capital (RAROC), which Connelly considers to be impractical. (Simply put, RAROC is net income divided by risk-adjusted capital.)

“A lot of the big banks like RAROC because they think it makes them more sophisticated, although they don’t do anything with the RAROC data,” Connelly maintains. “They don’t use it to risk price, they just make themselves feel good, saying ‘My RAROC is through the roof’.”

Connelly previously worked for a top ten bank. Speaking of his experience there, he says, “I saw all these deals coming through the credit committee with RAROCs that were super high. Yet, when I looked at our financial statements, [the numbers] didn’t agree.”

Not everyone agrees with Connelly’s take on RAROC’s value. Fintech vendor Syntellis, for example, advises banks to adopt a RAROC model, because it can mitigate the risk of a loan or series of loans regardless of the economic market.

“While profitability can be viewed through several lenses, RAROC should always be one of the calculation methods used by banks and credit unions,” Syntellis writes in a blog post. Other models can calculate if a portfolio is profitable, but it will not explain why the portfolio is profitable.

There are two other models nCino supports in its loan platform, Connelly says. One is net operating profit after tax, also known as a NOPAT. The one that Huntingdon is using, however, is net present value (NPV), which is calculated using customer acquisition cost and customer lifetime value (CLV).

Rethink Your Bank’s Performance Metric:

Evaluating profitability through RAROC might be the way the big banks do it, but community and regional banks could benefit from net present value.

“If my customer lifetime value is below my customer acquisition cost, I have a flawed business,” says Connelly. “I can’t make it up on volume.” The problem is that most banks typically don’t want to compile and analyze that data, he adds.

Read More: Why Data and Artificial Intelligence Will Transform the Future of Banking


In addition to profitability, Huntingdon’s new lending platform addresses risk issues using AI. Small banks often struggle with evaluating the risk of business loans in particular. Some local financial institutions avoid small business loans altogether, because they don’t know if they will be profitable, Connelly says. He says Huntingdon Valley Bank struggled with data analysis and cost structures before the nCino partnership.

“Now, we have three flavors of customer acquisition cost: small, medium and large,” Connelly says. He considers a small loan to be between $0 and $250,000, a medium loan between $250,000 and $2 million and a large loan to be north of $2 million. “I can now demonstrate that I can make money on a $50,000 small business loan the same way I can make money on a $5 million loan.”

Connelly says one of the artificial intelligence features they are most excited about from nCino is what the fintech coins nCino IQ or ‘nIQ’. If a lender is struggling to make a loan work for both parties, the vision for this feature is to suggest other products and services the lender may not have considered. For example, it might prompt the banker to offer a potential client a cash management package in addition to the loan.

This article was originally published on . All content © 2024 by The Financial Brand and may not be reproduced by any means without permission.