Modelshop Whitepaper

Innovation and Equity in Next-Generation AI Credit Risk Models

Enhancing the Credit Union Member Lending Experience with Real-Time Risk Decisioning Models

Executive Summary

Building upon the “people helping people” credit union philosophy, C-level decision-makers must continually rethink the member experience, especially as it relates to adopting smarter credit risk models that expertly harvest alternative data sources that create new member opportunities and enable frictionless transactions.

Gone are the days of a credit union relying solely on traditional risk models, such as FICO that founded its scoring model in 1956 and still focuses on payment history, amounts owed, new credit, length of credit history and credit mix. While a FICO score still has merit, with time comes change. In today’s marketplace, whether dealing with prime or subprime lenders, a credit union that relies on traditional credit scores alone is unwillingly experiencing a lack of precision, which leads to in consistent credit decisions on behalf of their valued members.

Credit union members are rightly demanding the same level of service from their primary financial institution that they receive from preferred online retail and ride share services, such as Amazon and Uber. So, credit unions, in turn, must provide a next generation experience powered by artificial intelligence (AI) that can interact with members, analyze risk and make data-driven decisions in real-time.

By better aligning risk and IT teams, credit unions can leverage tools which allow the development and deployment of modern AI-based risk models without a significant technology lift. This strategy uses no-code technology to connect knowledge-based employees and advocates into a next generation technology stack through application program interfaces (APIs). Taking a plug-and-play risk model approach reduces friction between risk analysts and the credit union’s IT department (or one that is outsourced). As such, credit union risk analysts can seamlessly pull data from the production environment and strategically design and test real-time decision models that can be reintegrated into the technology stack without the need to code or recode.

The purpose of this white paper is to educate credit union executives on the benefits of using AI to build, validate, deploy and monitor risk models that optimize real-time lending decisions and enhance member engagement and experience. To this end, C-level executives will better understand the impacts no-code, real-time risk decision models have on the credit union lending space, such as with credit cards, automotive loans and a home equity line of credit (HELOC), in 2024 and beyond.

The Drivers: Why Adopt a Modern AI Risk Decisioning Model?

In January 2024, news spread about an unintentional credit decisioning bias by one of the industry’s largest credit unions. This instance highlighted the need for every credit union, no matter its size or asset class, to re-evaluate risk models with an eye toward inclusion and fairness.

When a credit union confronts disparity across socioeconomic classes and demographics as to how lending decisions are made, the answer isn’t to place a thumb on the credit risk scale (e.g., affirmative action programs), but rather, the correct response is to build a better scale. Antiquated decisioning models are being replaced by new model sets that address ever-changing, real-time economic and financial data points related to each individual member. Credit unions, therefore, need better flexibility in how they define their lending criteria, especially when dealing with potential new members who have shorter credit histories.

According to the National Credit Union Administration (NCUA), as of September 2023, there were 4,645 federally insured credit unions supporting more than 138 million members. The number of credit unions with a low-income designation was 2,575, which is a critical distinction as many of these members are often unreasonably caught in the traditional credit risk scoring trap. In 2023, for example, total outstanding loans for all credit unions were $132 billion, which include residential property loans ($60.8 billion), auto loans ($29 billion) and credit card balances ($8.8 billion). If AI-driven credit risk decision models are deployed, credit unions could confidently lend more money to members (or potential members) once deemed unqualified by outmoded credit risk scoring models. This represents a win-win for both the member and the credit union.

A recent Ernst & Young (EY) report, “How SME lenders can build next-generation credit decisioning,” also found that credit unions, banks and other lenders are actively transforming respective Small and Medium-sized Enterprises (SMEs) lending business platforms. “[This is due] to mounting competitive pressures, including from incumbent market players and fresh entrants, heightened customer expectations and unprecedented opportunities presented by advancements in data technology and analytics,” the report noted.

Building on this premise, five driving forces behind credit transformation were cited: emergence of fintech and big tech lenders; changing expectations; richer, more accessible data; technology is better and cheaper; and heightened regulatory scrutiny. “A global EY survey of 6,000 SMEs shows the growth in those seeking alternative financial service providers, with 31% of SMEs indicating they would consider a fintech provider and 26% a big tech provider as a source of financial means,” the report offered.

Additionally, EY found that 48% of SMEs are interested in faster credit as a service and that “Scalable and efficient technology solutions are driving extensive digitization and automation across customer journeys, delivering enhanced support for customer interaction through AI assistants.”

Credit unions that leverage next-generation AI risk models to drive this new business performance model can implement applications for lower cost without undertaking major development projects. This strategy, which brings together all credit union departments in a unified member-risk vision, is squarely aimed at reaching more members and enhancing respective banking and lending experiences. To this end, modern risk models allow for a comprehensive use of data and multiple instruments decisioning, which provides a holistic view of the member relationship. This includes onboarding and validating new members (e.g., know your customer (KYC), identity risk, etc.), originating new products and optimizing member servicing.

Once the new modern model is implemented, credit unions can expect a more diverse membership base, higher performing loans, more competitive loan/credit offers and better compliance visibility.

Historic Challenges: Innovation within the Decisioning-Model

Credit unions, like all lenders, base a credit decision on the probability that the loan holder will repay the debt. Traditional credit models, however, do not accurately reflect today’s member’s complete financial picture. Classic model factors, such as debt-to-income ratio or the level of revolving credit utilization, are becoming less predictive of a member’s ability or desire to repay their debt. There are multiple additional factors impacting the financial health of today’s credit union member. To this end, excessive student debt, delays buying cars and homes, less loyalty to financial institutions and economic impacts of the COVID-19 pandemic have made it harder to accurately assess loan applicants using traditional factors.

As a result, many small-to-medium sized challenger banks utilizing modern-day financial tools and technologies have successfully entered the lending space. When considering the state of older credit risk models, along with present day competition, credit union decision-makers can’t be fast followers. In order for credit unions to retain existing membership, attract new members, and protect their reputation, innovation is required.

Credit unions, therefore, must figure a way to remove friction between its risk teams and the IT department. Historic pain points include incompatibility between the tools analysts’ currently use, such as Excel and Python, and the environments in which the credit union’s production system operates (i.e., hurdles related to legacy core technologies).

In the same manner these noted technologies have aged, so has the age of the average credit union member, which the NCUA noted was 47-years-old in 2023. While it is imperative to meet or exceed the banking expectations of this age group, credit unions must also remain hyper focused on the 18-to-34 age segment because according to the latest Census Bureau statistics, millennials and those younger represent half of the U.S. population (335 million people). Further, recent Gallup research found that millennials, at 27%, are the least engaged banking segment compared to Gen X (34%), boomers (38%), and the Silent Generation (46%).

A 2023 Consumer Pulse Study by TransUnion also found that only 54% of consumers reported having sufficient access to credit. “The gap between credit importance and adequate access was greatest among the youngest generation. Gen Z reported credit importance at 98% — followed by 96% of millennials, 91% of Gen X and 74% of baby boomers,” the report noted. “Yet, only 35% of Gen Z agreed they have sufficient access to credit compared to 54% of millennials, 51% of Gen X, and 70% of baby boomers.”

Emerging competitors understand these banking trends and, in turn, have proactively developed tools to make decisions on credit applications in a fraction of the time compared to traditional credit scoring models. Additionally, these fintech challengers are reaching a broader demographic of applicants due to offering ease-of-use apps. As a result, many credit unions are losing the confidence of existing members and potentially losing the ability to attract the next generation.

AI risk decisioning models, however, have the potential to level the playing field. A credit union, for instance, could provide real-time credit pricing decisions, while a customer is visiting a partnering website or recommend an investment portfolio in real-time as a member is adjusting their profile online. This type of advanced risk decisioning model could also flag possible compliance violations in online trading or more simply approve a member for car loan or mortgage while allowing interactive loan restructuring, all in real-time.

The Future: Deploying AI-Driven Member-Centric Decisioning

While it can be argued that certain technology advancements has historically led to a more depersonalized banking experience (e.g., call center bots, ATMs), with AI and central decisioning engines, a credit union now, more than ever before, can understand their members and build unique, personalized relationships using AI. When AI models communicate in real-time and are able to react to new data while the member is engaged, the result is improved consistency and a better one-to-one personal member experience.

By partnering with a trusted and experienced third-party platform that deploys AI risk-decision models as pluggable services, a credit union can more easily modernize existing platforms, including integrating risk analysts, product owners, and IT onto a single-risk automation framework. It’s imperative, however, that from the start these respected professionals, along with the selected third party, work as a team. This diligent approach optimizes the AI-driven credit risk model, which is based on the credit union’s unique member goals.

Completely outsourcing AI risk models, however, can lead to lack of control, reduced member personalization and reduced portfolio performance. Therefore, the collaborative risk model automation platform should provide easy-to-use, no-code technology that allows various stakeholders to work directly with new data sources, variables, business rules and advanced analytics. This proven approach ensures that every team member has instant access to data, logic and insights without the need for the credit union to undertake protracted coding, which allows for cross-departmental agility and stakeholder accountability.

To be clear, modern AI credit risk-decision engines do not totally discount traditional credit scoring metrics, such as for risk and fraud. The scope of a next generation, modern risk decisioning model also includes the application intent, pricing elasticity and conversion probability modeling.

These next-generation AI platforms provide a seamless data integration hub that will support plug-and-play connectors to a wide range of vendors as well as provide out-of-the-box logic and analytics that allow credit unions to leverage multiple alternate data sources in their decisioning.

For credit union members looking to apply for or obtain a loan in today’s marketplace, for example, there are four variables that best define “modern” lending: improved member experience; increased efficiency; competitive advantage; and reduced risk.

By automating and simplifying the loan origination process using AI, lenders can process loan applications more quickly, offering a frictionless loan origination process, which is a key differentiator for credit unions looking to attract and retain a new generation of members. This origination process also reduces the risk of fraud and errors, as automated systems can detect and flag potential issues more quickly and accurately than manual processes.

The manner in which “risk indicators” are viewed by lenders, both prime and subprime, can also be positively impacted by deploying AI-decisioning. These modern risk models, for example, aren’t simply like traditional scores with broad cut-off mechanisms; rather they simulate the projected financial performance of a credit applicant in real-time, which turns under-served but potentially high-performing applicants from a decline decision to an approval decision, representing a more equitable credit union lending policy.

The time and effort required deploying next generation, AI-driven credit risk models depends on a few variables. A credit union with assets exceeding one billion that has a larger IT and risk team (i.e., more than 20 employees) can expect a six to nine month journey, whereas a credit union with less than $500 million in assets and a smaller IT and risk team (i.e., less than 10 employees) can outsource the project and expect to go live in less than three months.

Even for smaller credit unions looking to outsource the technical implementation of its AI risk models, it is essential that the risk team, even if one person, is actively involved in designing their next-generation models. Modern AI risk modeling platforms make it easy to extract critical risk model controls and performance analysis tools into customized risk portals that allow risk executives the visibility and control needed to optimize their member experience and portfolio performance.

The Takeaway: How to Get Started?

While credit union executives may understand the need to update respective credit risk lending models, they are often faced with two critical variables: cost and required staff. Utilizing modern credit scoring models, however, doesn’t require a robust technology department and budget. Not every credit union, for instance, has a risk team and data scientists ready to start working with alternate data sources that could deliver a modern AI risk model. And even if they do have some of these skillsets, they often don’t have the tools to aggregate this data into meaningful predictors without writing custom code. The prospect of getting a new risk model through stringent compliance hurdles can also be daunting.

So sure, the idea of adopting a next generation AI-driven model could seem intimidating, but with the right technology partner, modern risk models can be easily designed and integrated into any existing credit union tech-stack in a manner of months.

Experts in this niche field recommend starting the process slowly and methodically, as opposed to a “rip and replace” approach. Rather, it’s best to integrate modern AI tools that blend new alternate data modeling tools with existing data and advanced analytics techniques as “intelligent add-ons” to existing mature systems and processes. To ensure a successful model rollout, credit unions should test, deploy and measure the performance of these new AI models by engaging in a “proof of value” exercise before launching a full suite of services.

There are, of course, costs involved with rolling out any new technology. Adoption of no-code, credit risk decision models, however, allows credit unions to reduce certain IT and compliance costs, so that they can refocus strategies on other areas of the member experience. To this end, by placing the responsibility of building and validating new risk models in the hands of the risk team, the credit union will greatly reduce costly re-coding, testing and error correction cycles. As a result, credit unions can better react and respond to market conditions, all without worrying about deployment errors or timing constraints related to coding cycles.

To remain competitive in an ever-evolving, tech-driven lending market, credit union executives should explore the benefits of using a no-code, AI-driven risk modeling platform to build, validate, deploy and monitor risk models that optimize real-time lending decisions and enhance member engagement and experience.

Making more equitable credit decisions based on a member’s total income, credit history, recent transaction history (e.g., utility bills) and work experience creates lending relationships not based on outdated demographics and credit scores, but rather on real-time financial performance metrics. This approach is positively impacting the lives of countless consumers across the nation and credit union members are taking notice, as should credit union decision makers.



You cannot copy content of this page