NAFCU Services Blog

May 03, 2024

Ensuring Safety and Soundness with AI

By: Allison Huber, Head of Content Marketing at Upstart

AI and machine learning has firmly planted its roots in today’s financial landscape, signaling a paradigm shift that credit unions are eager to navigate. Yet, in this journey toward embracing technology, concerns loom large – cost implications, the learning curve and the ever-important task of compliance management. 

Credit unions must have a firm understanding of AI compliance and regulations to effectively steer through potential risks on the horizon. To make safety and soundness even more top of mind, in late 2023, President Joseph Biden issued an Executive Order to manage AI risks while also promoting innovation. It included new standards around AI safety and security such as new safety testing, the establishment of new standards and best practices for authenticating AI content and the establishment of new cybersecurity programs to help fix vulnerabilities in software.1 

Lynn Sautter Beal, Head of Implementation and Success at Upstart, joined the latest CU Lab podcast to address common concerns credit unions have with AI and how they can have confidence while walking a regulator through AI models. Sautter Beal brings a wealth of insight garnered from her extensive tenure in banking. Beal said regulators are more focused now on examining fintech relationships. With the financial services industry growing its use of AI, regulators are looking at fairness in lending especially when AI is being used with outside models. 

Find the Right AI Partner 

Beal advised credit union leaders to prioritize compliance expertise when considering a fintech partnership: “Look for depth of experience, expect transparency and ask how they evaluate new data sources,” said Beal. Also, when it comes to compliance, determine how the organization continues to evaluate, test and make improvements over time. Upstart, for example, has demonstrated its commitment to compliance as a founding member of the American Fintech Council; the company is committed to promoting a transparent, inclusive, and customer-centric financial system along with pioneering products to better serve the underserved.2 

Prepare for AI Compliance Risks 

A key component is designing an appropriate governance framework. One tactical step credit unions can take to manage compliance is identifying where AI cannot replace humans since AI systems can sometimes lack contextual understanding. The first line of defense against discriminatory AI should include some form of manual review. Also, ensure that AI uses are properly vetted before launching into widespread adoption. Credit unions understand that while they may gain in increased productivity and cost savings, if something goes wrong, their reputation is on the line.3 

AI Expertise is a Must 

Credit union leaders should partner closely with their fintech partners, especially to overcome objections and questions from their team. “If you are the champion and you know there are objectors, talk to your account executive at your fintech and have them help you overcome those objections,” said Beal. She encourages credit unions to ask if risk and compliance truly have a seat at the table. Also ask, is there a true partnership between the risk and compliance teams, the products teams, and the machine learning or AI teams?  

Beal added that fintechs should have case studies showing how they have built out new members, as well as how they have innovated cross-sell and deepened relationships. Beal added that credit unions should ask about core competencies, including evaluating expertise in their management team. 

Prepare to Partner 

Beal said that fintechs should build a diligence space that will “educate and empower credit union partners.” That includes helping credit unions present the business use case to their Board. She stated that Upstart supports having that successful conversation by articulating the business use case to the credit union board. Upstart’s Regulatory Readiness program includes continuing education and overcoming obstacles such as having enough resources to provide oversight, what happens after an exam, and how to talk about AI when credit unions may not have internal expertise. “Lean on your fintech partner to support you through challenges, build your program, and answer your objectives,” explained Beal. She added that Upstart has the track record to show its partnership “really helps you build that business” after launch.  

Partner with The Leaders in AI

Beal’s advice to seek deep experience has brought many credit unions to Upstart, as a leading AI lending platform. This fintech can help credit unions approve more borrowers safely while growing membership. “It’s not set-in-and-forget-it,” said Beal, adding that Upstart helps credit unions with training to prepare for regular reviews. The all-digital experience that originates loans in minutes results in over 87 percent of loans funded are fully automated with no human interaction. Beal said when credit unions have a handle on reserves and capacity for lending, they can start small and grow. 

Upstart’s AI model leverages over 1,000 data points and advanced machine learning algorithms to enable more accurate risk-based pricing. With Upstart, credit unions can approve over 44 percent more borrowers than traditional credit score-based models with low default rates.4 Upstart works side-by-side with credit union teams to ensure they have full visibility and control over their lending program. 

  4. As of October 2023, and based on a comparison between the Upstart model and a hypothetical traditional model. For more information on the methodology behind this study, please see Upstart’s Annual Access to Credit results here

About the Author