4 Ways Automation Can Help Your Credit Union

As contrary as it may seem, using AI in your underwriting process can actually make the experience more human for your members. It’s faster, more accurate, helps you maintain compliance, and lets you approve more borrowers, while improving the experience they have with your credit union.

From broadening access to credit, to making the underwriting process more consistent, here are four ways automated underwriting can help your credit union. 

1. Broaden access to credit 

At Zest AI, our mission is to make fair and transparent credit available to everyone. In the United States, approximately 47 million people are left out of the credit system because they are difficult and often impossible for traditional credit scores to assess.  That’s because credit scores use just a few dozen variables to determine a borrower’s creditworthiness. This  disproportionately affects immigrants, people of color, and low-income Americans because the limited set of variables used in credit scores doesn’t effectively describe them. 

That barrier to credit doesn’t impact just the individual borrower — it affects their families and the generations after. According to the Urban Institute, young adults of color, particularly those in majority-Black or Hispanic communities, start adulthood with lower credit scores than their white counterparts. Decades of discriminatory practices in the American credit system are reflected in credit scores, which continue to perpetuate inequities in racial and generational wealth.

The solution to closing that gap is multi-pronged, and responsibly increasing access to credit is a key part of the solution. The legacy credit scoring system is broken, but AI-driven lending can help to fix it. By using more data, better math (i.e. machine learning), and adopting less discriminatory models, we can make strides in creating a fairer system. For example, on average, lenders that switch from using credit scores to Zest models have been able to increase approvals for women and people of color by 30-40% with no increase in repayment risk.

AI also makes it possible to automate more loan decisions, which not only speeds things up significantly but also makes underwriting more accurate and more consistent. This allows you to confidently automate approvals and say yes to more members - including those who would have been excluded by legacy credit scores. ‍

2. Get a higher fidelity view of your borrower

Imagine if you expanded your assessment of creditworthiness to consider several hundred data points rather than just a few dozen. You would get a much clearer picture of your borrower and their creditworthiness. 

Credit scores provide a very un-nuanced view of the borrower using limited data.  According to Business Insider, 65% of the U.S. credit scoring model is based on credit utilization and repayment history. Of course, credit utilization has to do with how much credit you were given to begin with. Since many people of color have lower credit scores, they are given lower credit limits, and so their utilization will naturally be higher than their white counterparts. This further perpetuates lower scores.  

Socio-economic differences do exist between groups, and these differences can cause a legitimate difference in repayment risk.  But we need to be as fair as possible given our history of discrimination and bias.  The main issue today with the U.S. credit scoring system lies in the methods credit score providers use to calculate their proprietary, black-box scores. With AI-enabled credit models, we can incorporate more data points to assemble a more complete picture of borrower risk, and we can use better math to more accurately assess risk.  Unlike traditional credit scores, the AI models Zest builds are completely transparent so you can audit and inspect them.  AI models allow you to  lend to borrowers who might have otherwise been overlooked. If we do that, we can bring over 40 million Americans back into the credit system.

3. Make auto-decisioning inclusive and consistent

Although it is unlawful to discriminate in lending practices, the problem still persists. Human beings can be biased and are subject to mistakes or inconsistencies. Imagine your loan officers are having a particularly busy day with increased volume - perhaps the same logic and rigor is not applied from one application to the next. Things like that can lead to inconsistencies in the underwriting process.

It has been known for some time that automation can help remove unconscious biases from the lending process and improve access to credit. For example, a 2003 report by Susan Gates, Cindy Waldron and Peter Zorn showed there were higher mortgage application acceptance rates for low- and moderate-income households and neighborhoods, including people of color, when using automated underwriting.  Automated underwriting was first introduced by the GSEs in 1995 and was met with great skepticism, yet it proved to improve access to homeownership for millions of Americans. Greater automation leads to greater inclusivity, but its positive effects can be even greater when the model used to automate the underwriting process is more inclusive.  

Machine learning, and the associated increase in automated underwriting it makes possible to ensure your lending decisions are consistent and accurate 100% of the time. However, we know that human-made artificial intelligence is not immune from bias or fallacies. That’s why it’s important to be purposeful in how you build your AI models. You want to ensure that you’re eliminating bias wherever you can so that you get a fair and consistent outcome.

4. Make your members happier 

As we talked about in our last post, Take Back Your Time with Auto-Decisioning, integrating AI into your underwriting process can make your members happier, because it increases the number of loans that can be auto-decisioned. You don’t want a member to wait around for a decision — wait too long, and they might go someplace else. Because machine learning models are more accurate, they make it possible to auto-decision more loans, so you can get a potential borrower an answer instantly. 

The average credit union reports that it takes over 30 minutes for a borrower to get a decision on a loan. In the era of same-day delivery and instant gratification, that just doesn’t fly. By automating your underwriting process, you can have a decision within seconds — not half an hour. Not only are your members happier because they had a better experience, but you’ve also given them back their time, as well as your own, to focus on the things that matter most. 

Think of the myriad of ways AI and automation have already been incorporated into your everyday life: Google search, the ATM you use to get some cash, Netflix recommendations. We don’t necessarily notice these technologies as we use them, but we do notice when we don’t have them. Think about how much time it used to take to look something up manually instead of using Google, or the line at the branch we used to wait in to withdraw money, or the time spent at the video store in search for the perfect show to watch after dinner. Would you want to go back to the old way now that you know the alternative? Those extra seconds, minutes, maybe even hours, can drive significant changes in consumer behavior — consumers that credit unions want to serve before they turn to a more convenient solution.

Automation and AI have forever changed how we do many things, and it is time to adopt these technologies for lending so borrowers can get the most accurate and fair decision quickly.  AI-driven underwriting can help you deliver this better and have a more humane member experience.