The changing economic environment is opening up opportunities for banks and credit unions to implement artificial intelligence models in loan underwriting throughout the credit cycle.
The evolution of digital banking over the past decade has provided much richer data on consumer profiles and behaviors. The company has cultivated a group of financial technology vendors to develop his AI underwriting model for banks and credit unions. Since most of them were established in a favorable environment, early adopters are now looking to validate their AI models in the current context, where downside risks leading to credit deterioration are high.
VyStar CU Vice President of Consumer Lending Alice Stevens
Source: VyStar CU
Alice Stevens, vice president of consumer finance at VyStar CU, said: “We will be closely monitoring this model to see if it is working as expected during a time of economic change.” .
VyStar's AI journey began about four years ago. In 2019, the Jacksonville, Florida-based credit union partnered with ZestFinance Inc., doing business as Zest AI, a fintech company that builds AI models for lenders. In 2021, VyStar and First National Bank of Omaha led his $18 million investment in Zest AI.
Founded in 2009, Zest AI tested its model using data in a 2008 economic downturn scenario. But for VyStar, it's the first time to actually see how the model performs when delinquency rates are rising and consumer attitudes are changing due to inflation. Mr. Stevens said of the cash flow pattern:
“I believe in this technology, but I'm always going to be cautious about how far I jump into the water before it feels warm,” Stevens said.
Pagaya CEO Gal Krubiner said the use of AI models in credit underwriting was led by digital lenders such as LendingClub. Krubiner said that as digital banking becomes more pervasive in the operations of financial institutions, a key factor driving the use of algorithms is the generation of clean and standardized data. Founded in 2016, Paaya provides a network of AI models to make credit decisions for financial institutions such as Upgrade Inc., Ally Financial Inc., and SoFi Technologies Inc.
“I think a much wider range of financial institutions are beginning to develop and use machine learning models for credit purposes than they were three or four years ago,” said Tori Shinohara, partner at Mayer Brown LLP.
conservative approach
Unlike technology companies that aim to make machines smarter, banks and credit unions are taking a more conservative approach that relies on human assistance.
Zest AI uses supervised machine learning, a subcategory of AI, to help banks create records to explain credit decisions. In a supervised AI model, lenders use appropriately labeled input and output data to train a machine to predict outcomes.
“In our model, we can tell you what the variables are and we actually provide that documentation,” said Yolanda McGill, vice president of policy and government at Zest AI.
By comparison, unsupervised models do not necessarily require human intervention. But industry advisors say the reliance on algorithms itself makes it difficult to visualize the basis for credit decisions.
“What I'm seeing being adopted today, especially in regulated institutions, is going to be supervised machine learning models,” Mayer Brown's Shinohara said. “So it's not a complete black box. To meet the explainability requirements under the law, we need to know what's in and what's coming out.”
In guidance issued Sept. 19, the Consumer Financial Protection Bureau said lenders that use AI or other complex models must specifically He emphasized the need to provide accurate and accurate reasons.
VyStar uses AI models to scrutinize loan applications for large volumes of consumer financial products.
Source: VyStar CU
Because VyStar uses a supervised model, loan officers play a key role in monitoring performance and conducting tests to uncover potential errors. VyStar primarily uses AI models to vet loan applications, and has applied them to most consumer financial products, including credit cards, personal lines of credit, and auto loans, Stevens said. Stated. He said more than 60% of loans using AI can be approved instantly, while traditional digital lending solutions can only approve about 30% instantly.
AI models are better at building correlations between data points to understand an applicant's creditworthiness, rather than simply checking “yes” or “no” like traditional digital lending software. Stevens explained. The credit union is also interested in exploring the use of AI in more complex tasks, such as loan pricing, she said.
“because [AI] Using data in a relative way makes it more reliable,” Stevens said.
Regulatory notice
After the Great Recession of 2008, fintech innovation flourished, due in part to low interest rates and investor interest in the sector. But regulators have long been wary of a lack of evidence of fintech companies' ability to survive through cycles.
In its 2017 report, the Financial Stability Board wrote:[W]Without the benefit of a full credit cycle, it is too early to say how new models that leverage big data will perform in terms of risk measurement and pricing. ”
But innovation is happening. ChatGPT has taken the internet by storm since his November 2022 announcement, sparking widespread discussion about increased regulatory oversight of AI. In particular, financial regulators have emphasized that AI models comply with existing consumer protection laws.
Sen. Elizabeth Warren (D-Mass.): “AI is getting a lot of attention in Washington. Big tech is coming to tell us how to shape new laws that advance their business models. But , laws already exist to govern some aspects of AI,'' he said at a Senate Banking Committee hearing on September 20.
Like other lending practices, the use of AI tools for credit is regulated by the Equal Credit Opportunity Act and the Fair Credit Reporting Act. Zest AI's McGill said that when it comes to AI models, what banking regulators are looking at is the ability of lenders to explain credit decisions and fair lending.
Pagaya's Kulbiner added that regulated entities are also expected to have a governance framework around algorithmic models. Pagaya has been working with his associates Deloitte LLP and Charles River to audit AI models for credit decisions.
“There is no 'AI exception' to our consumer protection laws,” Warren said at the hearing.
VyStar will continue to monitor and optimize its AI systems to ensure they comply with regulations. If this approach is successful, the company expects to improve efficiency.
“Eventually there will be fewer human underwriters because they won't be reviewing as many loans, but that will all happen over a long period of time, a fairly long business cycle,” Stevens said. Told.
