AI opportunities

 

 

Your best application for artificial intelligence (AI) might be finding opportunities you never knew existed.

Credit unions are exploring AI-based technology solutions to:

• Identify worthy loan applicants who might otherwise face rejection.
• Boost loan volume with automated decisioning.
• Improve security through voice recognition.
• Integrate voice, video, and smart speaker channels.
• Predict potential risk within loan portfolios.
• Explore the possibilities hidden within their data.

The projects are varied, but they share a common goal: to go beyond business as usual while gaining the AI experience required to remain competitive.

Getting started

AI is a branch of computer science that blends computers’ data capabilities with human-like intelligence to complete specific tasks. AI solutions often include “machine learning” algorithms to capture specific data from interacting with human users, which enables an AI tool to improve over time.

John Best, CEO of Best Innovation Group, advises credit unions to take advantage of their “superpower of collaboration” to explore AI through credit union service organizations and joint projects with financial technology (fintech) companies.

He says credit unions should approach AI with the same intensity allocated to a core processing conversion. The first step is strengthening data analytics and developing data leaders.

“If your data analytics people don’t already have a seat at the table, you’re behind,” Best says. “This is a culture shift in the organization.”

He adds that credit unions must also think about what’s possible with AI.

Increasing loan acceptance

Northern Hills Federal Credit Union in Sturgis, S.D., began using an AI-based loan decisioning tool to increase approved loan applications by a projected 25% to 40%.

Floyd Rummel III, CEO at the $125 million asset credit union, says members used federal stimulus payments to increase savings and pay off loans during the pandemic. That dropped Northern Hills Federal’s loan-to-share ratio from 78% in 2020 to 68% in June 2021 and reduced interest income.

The credit union uses a loan decisioning tool to reverse that trend by sorting applications into three categories:

1. Green applications qualify for automatic approval based on credit score so members can get quick, after-hours approval online.

2. Red applications fail to meet lending standards and are declined with an offer of help and information. 

3. Yellow applications don’t qualify for automatic approval but might qualify after the credit union gathers more information. The tool relies on LexisNexis data to add information about utility bill payments, address changes, and other factors.

Previously repaying a credit union loan also impacts decisions.

Rummel says Northern Hills Federal focuses on “yellow” applications with the goal of approving more loans more quickly. As the credit union reviews and eventually approves or declines these applications, the AI tool learns how to handle similar applications in the future and place more of them in the right category without employee intervention.

“As the yellow category becomes smaller, my loan people have additional time to spend with these applicants,” Rummel says.

Northern Hills Federal prepared for launch by:

• Compiling five years of loan application data through the AI tool to confirm it’s reviewing loans appropriately.
• Adjusting parameters for acceptable levels of charge-offs and delinquencies, which are expected to increase along with loan volume.
• Reassuring employees that AI will reallocate resources, not eliminate jobs.

Rummel says the biggest challenge was integrating the AI tool with the online service bureau the credit union uses for core processing. He hopes Northern Hills Federal’s efforts will prompt other small credit unions using the same service bureau to pursue AI.

NEXT: Reducing call center fraud

Reducing call center fraud

TruWest Credit Union in Scottsdale, Ariz., began using AI in January 2021 to offer biometric authentication for members who contact the call center.

Chris Kearney, chief information officer for the $1.4 billion asset credit union, says call center volume was rising even before the pandemic pushed it higher, making it important to handle calls efficiently. 

Traditional authentication is based on a time-consuming process of gathering answers to a list of personal questions such as “mother’s maiden name.” Fraudsters use dark web data to answer those questions and gain access to account information.

“We were actively looking for a solution to add to call center agents’ toolbox to streamline interactions and protect members’ identities and funds,” Kearney says.

TruWest rejected existing voice recognition solutions as too expensive and cumbersome until it found a service-based AI tool from Illuma Labs.

“Our voice authentication solution uses AI to analyze member calls in real time, matching the results against an enrolled member’s voice print,” Kearney says.

The 15-second process replaces a typical 90-second question-and-answer authentication session.

Machine learning is built into the solution so the voice signature gets stronger every time a participating member calls the credit union. 

TruWest launched voice recognition with a three-step process:

1. Define risk tolerance. A cross-functional team drawn from the call center, information technology, risk management, employee development, and leadership weighed AI’s ability to reduce call center risk. A call center team tested the system with a pilot in May 2020 and a soft launch in November 2020.
2. Offer privacy. Call center staff developed a narrative to explain the benefits of voice recognition and offer enrollment to members. Almost 10,000 members—nearly 11%—use voice recognition as of June 2021.
3. Update policies and procedures. The credit union updated its privacy policy to address biometrics before voice recognition began, for example, and confirm members’ ability to opt out.

TruWest is also using voice recognition for outbound collection calls and will soon apply it to authenticate online chatbot users. The online chatbot relies on AI to automate the exchange of information with members online.

NEXT: A collaborative approach

A collaborative approach

Collaboration allowed BCU in Vernon Hills, Ill., to cost-effectively explore using an AI-based conversational chatbot to interact with members.

The $4.9 billion asset credit union worked with $4.4 billion asset STCU, Liberty Lake, Wash., and $3.5 billion asset Connexus Credit Union, Wausau, Wis., to explore the “virtual financial assistant” offered by Abe.ai in 2017. 

“It’s like when you’re ready to dive off a cliff you gain confidence when you have partners standing beside you,” says BCU Senior Vice President Carey Price.

BCU introduced its chatbot in 2020 to offer “conversational banking” to members who use text messaging services and smart speakers such as the Amazon Echo. 

The credit union upgraded its video banking to enhance its appeal for members who became accustomed to handling transactions without visiting a branch during the pandemic.

As a next step in its engagement strategy, BCU plans to use its conversational banking tool as the conduit for integrating data from video, telephone, smart speaker, text, and online channels.

The data will be funneled through Salesforce, which will act as the unified agent desktop (UAD) to track data about every interaction while enabling call center agents to use tools such as co-browsing to share screens with online members.

Price cites multiple benefits for credit unions that collaborate to explore AI:

• Sharing the load. Each credit union took on specific tasks and then shared the results and the costs.
• Breaking a path. Because the three credit unions use the same core processor, the core processor was more willing to invest in integrating the AI tool. That provides a pathway for other credit unions who rely on the same core processor to begin using AI.
• Attracting fintech partners. Fintechs in search of clients find three credit unions more appealing than a single organization.

“I encourage credit unions of all sizes to collaborate and lean in and explore,” Price says.

Exploring data

Suncoast Credit Union in Tampa, Fla., is using AI for exploratory data analysis. Data Analytics Director Nolan Walker says the result is “directional knowledge” about opportunities hidden within operational and member data.

The data team at the $14 billion asset credit union built a “supervised learning model” in 2020 to mitigate internal risk, for example. The model leverages machine learning to gather information from features within basic lending metrics such as delinquencies and charge-offs.

The model predicts the outcome of a given product for the following 90 days, including projected delinquencies, the likelihood a delinquent loan will become a charge-off, and the potential for a member with a delinquent loan to independently remedy the situation, which reduces net risk.

Walker notes that AI projects assigned to the data team make high demands on time and talent. When AI is not the most effective solution to the problem, this cost tends to outweigh the benefits.

Overcoming barriers

Walker highlights two common barriers to using AI to explore existing data:

1. Poor data quality. Credit unions often created or purchased legacy technology solutions to perform specific tasks without regard for whether the data they captured could later be aggregated or interpreted for other uses.
2. Inability to explain benefits. Data leaders may find it hard to explain AI goals to leaders and employees if they offer vague ideas, define key terms poorly, or fail to explain how member data will be insulated from front-line decisions about products and services.

AI often challenges traditional ways of doing business, which makes it harder for leaders to trust that it will be applied in members’ best interest. Walker says aligning AI with strategic and business goals is crucial, as is allotting time to creatively explore the technology. “The quicker you get this in front of your business leaders and then leverage the tough questions you get from them, the better off you’re going to be.”

NEXT: A disruptive force

A disruptive force

Kirk Drake, president and founder of CU2.0 and author of “FinAncIal: Helping Financial Services Executives Prepare for an Artificial World,” predicts AI could prove more disruptive than the internet as it changes “our entire relationship with technology.”

For example, Drake says a financial institution could use an AI-based tool to assemble a pattern based on how a specific member handles financial matters. The member could then opt into an AI-based, personalized tool that would use that pattern to automate the ongoing process of moving money and paying bills.

Drake says the nation’s largest banks likely have a five-year head start on exploring these types of “transformative” AI applications. 

That makes it more urgent for credit unions to take advantage of their ability to share data and services, as well as fintech’s need to find partners in delivering AI-based financial services.

“We’re the right ecosystem to experiment,” he says.