With the explosive buzz that artificial intelligence models have brought into the mainstream, AI, especially generative AI, has been almost impossible to ignore. But while ChatGPT and similar tools attract attention and drive excitement, banking companies have successfully deployed AI and machine learning models for years to propel growth, enhance efficiency, and gain a competitive edge on the lending side. That’s pretty exciting and has a track record.
AI presents many promising opportunities for banking institutions in lending, revolutionizing how financial institutions assess risk, underwrite loans and enhance customer experience.
Artificial Intelligence Augments the Human Element
AI is not meant to reduce human staff, nor is it meant to entirely supplant the human role in making credit decisions. No matter how sophisticated the technology, you will always need some layer of human touch.
The most effective applications are those that use AI as a supportive force: aiding in data analysis, helping to craft predictive modeling, and serving as a source of information for better decision making. This application is less glamorous than large language models and other generative AI, but operational optimization is one of the most tried and true applications for artificial intelligence. This optimization also ultimately unlocks new opportunities for upskilling and reskilling the institution’s workforce.
Not only can AI help banks become more operationally efficient, but also more secure. Take fraud detection: Traditional AI employs robust machine learning algorithms to analyze historical data, patterns and anomalies, enabling lenders to detect and prevent fraudulent activities in real-time, in an effort to minimize financial losses.
Over time, machine learning models get smarter and accelerate the accuracy of fraud detection. A key consideration here is data quality. The better the data that goes in, the more sophisticated and tailor-made a model can be.
Some financial institutions may not have the volume that provides a really robust database to build a strong fraud AI function on. This isn’t the end, however. A bank or credit union can work with an AI partner that serves multiple institutions. The models they build can draw on patterns seen among thousands of applications daily.
Read more: ChatGPT Will Become ‘ChatOMG!’ in 2024, Forrester Predicts
Using AI to Find More Business through Affiliate Websites
A common limitation of some models is that they are quite general. They don’t account for all the nuances and risk appetites of the particular financial institution using them. Well-applied AI, however, should enable lenders to build bespoke credit models tailored to specific customer segments or niche markets. These models consider unique data points and variables, enabling more accurate credit assessments and personalized lending decisions.
Wrestling for Affiliate Site Loans:
The importance of custom credit models is emphasized when a lender is operating within the affiliate marketing channel. This includes such sites as Credit Karma, LendingTree and Bankrate. Competition is intense and it's live.
Multiple lenders get a look at people who want particular types of loans. They have to offer the features the would-be borrower wants and the applicant, on the other hand, has to fit the institution’s credit profile.
Affiliate channels are particularly competitive, making it critical for lenders to have more accurate predictors of credit risk. Simple credit policies relying primarily on general credit scores like FICO will not capture the variance present within each FICO quality band. More sophisticated lenders will be able to target the 750 FICO customer who actually behaves like an 800, while less sophisticated lenders will only see the 750 FICO customers who behave like 700s.
It’s a matter of the differing approaches to marketing credit.
Want to go deep on AI best practices for banks?
Attend our AI Masterclass — Unlocking the Power of Artificial Intelligence in Banking — at The Financial Brand Forum 2024 on May 20-22 in Las Vegas. Led by Ron Shevlin, chief research officer at Cornerstone, this three-hour workshop will be jam-packed with lessons learned from industry leaders and real-world case studies.
For more information and to register, check out the Forum website.
Existing customers or local prospects who come directly to a banking institution tend to be easier to underwrite and are generally less price sensitive.
However, in a competitive marketplace like an affiliate site, consumers are far more likely to simply take the best offer. There’s no familiarity, no brand loyalty.
So then the competition comes down to the sophistication of the tools. If a financial institution relies on a single score, like a FICO or Vantage, then a competitor with more sophisticated credit models will be able to tweak what they offer more finely. This will enable them to offer the best customers the best rates, leaving the institution with the simpler approach — unknowingly — with a higher-risk population.
AI models that are customized to the lending institution lead to more accurate pricing. This can lead to higher conversions for superior borrowers, supporting higher growth, reduced operating expense, improved customer experience and, ultimately, higher profitability.
Read more:
- Notably Quotable: What Banking and Tech Leaders Think of Generative AI
- The Importance of Continuous AI Innovation in Banking
- Why AI Tech Projects Often Flop in Banking and What to Do
Meeting the Challenges of Launching AI-Backed Lending
Many banking institutions have embraced AI adoption and are well into implementation, but many others have only gone so far and others are still hanging back. One concern that slows adoption is regulators’ reaction to the institution’s adoption of the technology.
Maintaining regulatory compliance standards is always a priority. In speeches and panel discussions, regulators frequently mention their concerns about lending via AI — the possibility of models picking up or even developing biases. And there’s a certain distrust for “black box” decision making.
Another concern is talent, as data scientists aren’t cheap and it can be an expensive skillset to keep happy. And having the right data can be expensive. Beyond your own proprietary data sets, you need a standard of comparison against which to benchmark your own performance and customer base.
The prospect of integrating AI into lending programs may seem daunting for some banks. They must recognize that solutions and resources are readily available to address concerns and facilitate the institutions’ transition.
Ultimately, embracing AI for lending is not about achieving immediate perfection but rather about taking the first steps towards progress. It requires an upfront investment of time and financial resources, but the long-term benefits and dividends are seen in operational efficiencies, risk mitigation and enhanced customer experiences.
About the author
Garrett Laird is director of decision science at Amount.