The advent of generative artificial intelligence (GenAI) presents an extraordinary opportunity to reshape the dynamics of consumer loan underwriting. By facilitating an interactive, collaborative underwriting process, GenAI could empower consumers to engage directly with lenders, helping consumers understand how their personal financial data is used to determine the price and availability of loans they are being offered and to learn what factors or behaviors might produce better financial outcomes.
So how might this new GenAI collaborative underwriting model work? It would begin with the lender/underwriter collecting as much relevant data as it can regarding the consumer. The underwriter would create a digital financial profile for the consumer including both the affirmative and negative data points. Then the consumer would be offered an opportunity, using generative AI technology provided by the lender, to access his/her file electronically and to pose questions regarding underwriting results (and, of course, correct any inaccurate data or offer additional relevant information). The consumer might ask questions like: "If I close one or two of my credit card lines, how would that affect the amount I could borrow or the interest rate?" or, "What would happen if I made a larger down payment?" All consumers would be given the same transparent access to data and the right to ask questions using the GenAI tool provided by the lender, thus promoting equal and fair treatment for all borrowers. The consumer would be offered an opportunity to "pull back the underwriting curtain" and see how data regarding their financial profile is being used to predict their likelihood of success or default thus influencing the price and availability of the credit being offered. Disclosures would become dynamic rather than static.
And why should lenders or consumer advocates support this process? Well, as the 2008 financial crisis demonstrated, both the borrower and the lender have a strong interest in avoiding default. A loan that is too risky for the consumer is in no one's best interest, certainly not the consumer's. And a loan that is overpriced carries the risk for the lender of discrimination or abusiveness challenges … as well as the possibility of an expensive default by a borrower who overborrowed. If the lender and the borrower collaborate in finding a loan that best meets the borrower's needs, both parties benefit.
Of course, anyone who has used GenAI knows that, if it is not tethered to a reliable data set, it can provide "imaginative" and inaccurate answers. So, the first challenge to the lender/underwriter would be to make sure that the data set and the analytic tools that make up its underwriting model are fully vetted, as they should be anyway.
For more than 50 years, consumer loan underwriting has relied almost entirely on data provided through credit reporting agencies, related credit scoring systems like FICO and data produced by underwriting engines operated by large lenders, government-sponsored enterprises (GSEs) or more recently by fintech companies. Availability of credit and the cost of borrowing have been determined by results produced by these data engines whose workings are little understood by the average person.
Regulators remain skeptical toward digital assets in the wake of major crypto-industry partner-bank failures, but in the absence of legislative direction, oversight of crypto is largely left to the Securities and Exchange Commission. The SEC's approach has been aggressive, but its authority has been muddied by recent court decisions.
Two fifty-year-old consumer protection laws, the Equal Credit Opportunity Act (ECOA) and the Fair Credit Reporting Act (FCRA), require that consumers be given an opportunity to review the information in their credit files to determine accuracy and make any necessary corrections. Lenders must also provide the consumer with information regarding the factors that impacted the price or availability of credit if the consumer did not receive the best terms on offer. But these static, generic explanations tend not to be very satisfactory. More recently, state-enacted privacy laws have sought to provide consumers with access to personal information about them held by third parties, and the federal Gramm-Leach-Bliley Act gives consumers some power over who has access to their personal information. Finally, Section 1033 of the Dodd-Frank Act seeks to give consumers more access to and control over their financial data, and the CFPB is currently engaged in rulemaking regarding this law.
Collaborative underwriting using GenAI would not require an amendment to existing law or regulation. Nor would it make any less important the data collection and sharing functions performed by the CRAs and others. Current regulatory requirements would be complied with, and consumer access to information provided by the collaborative underwriting model would be an added benefit to the consumer offered by a lender only if it chooses. If GenAI is to be used to create a collaborative consumer underwriting experience, it will probably not be the result of a regulatory mandate, but rather as an initiative by some private sector player (a bank, a credit bureau, a tech company?) with Steve Jobs-like vision and commitment to enhancing the borrower's experience. Being a first mover is not for everyone, but a consistent commitment to enhancing the consumer experience can produce expanded market share and customer loyalty as Jobs demonstrated.
At a time when the media is filled with stories focused on the potential negative impact of GenAI, wouldn't it be helpful to demonstrate the consumer benefits that can be achieved using this technology? It is even possible to imagine a GenAI collaborative underwriting model of this kind being used in high school classrooms to help students understand how their future financial, employment and behavioral decisions might impact their access to credit.
When ECOA and FCRA were enacted in the 1970s, the huge computers at the Pentagon didn't have the computing power that a handheld has today, and advanced AI and large language models (LLMs) were distant dreams of science fiction writers. Today they are a reality. Can technology help achieve better consumer understanding and protections in ways that old laws really cannot? This suggestion for a new GenAI-enabled collaborative underwriting model using dynamic disclosures would not initially require amendments to ECOA or FCRA (overlapping compliance with those regulatory requirements would still be required). But if a better way of ensuring fair, effective and transparent use of consumer data for underwriting could be demonstrated, perhaps those laws might eventually be modernized.
