12 minutes
Credit unions’ famous hospitality expands with AI, chatbots and financial advice options.
"Alexa,” says the happy credit union member as they drive to the beach, “move $700 from savings to checking, pay all bills due by the 10th, and apply anything left to my credit card balance. And if I’m down to 10 checks or fewer, order some new ones. Oh, and set up any CDs that mature in the next 12 days to roll over into whatever CD is paying the highest rate for maturities under 25 months.”
Today’s credit union members aren’t quite that happy—yet—but that is a fair example of what they are coming to expect. And it’s what CUs are asking their vendors to help them deliver.
Practical applications of artificial intelligence remain elusive. A digital revolution is sweeping financial services but not yet showing up as productive AI at most CUs, reports Kirk Kordeleski, CCE, senior managing partner and chief strategy officer of BiG (Best Innovation Group) Consulting, Tampa, Florida.
“Digital has moved from piecemeal activity to something affecting everything credit unions do,” he notes. But AI depends on the convergence of digital communication, deep data collection and analysis and sophisticated programming—and CUs are struggling with the data piece, he points out.
With the proliferation of delivery channels, disparate systems and data pools have sprung up. “Most credit unions are struggling to integrate data, which is necessary before they can introduce machine learning and AI,” he says. “They need to resolve this issue first.” Fintechs and large financial institutions with data scientists on staff are reaching that point now, he adds.
Artificial intelligence for CUs is still a goal, not a reality, agrees Sabeh Samaha, president/CEO of Samaha & Associates, Chino Hills, California. However, weighing evidence to make informed decisions is certainly a big component of human intelligence, and that’s something CUs have already learned to automate in a limited but effective and productive way, he points out.
“Decisioning matrices are not new,” he points out. “In many cases, decisions to enroll a prospective member, provide an indirect auto loan, offer a credit card and set risk-savvy limits have been successfully turned over to technology and happen quickly and automatically.
“If you identify the criteria that people use to make such decisions, and if you collect and apply the data they would base those decisions on, you can automate decisions. And credit unions have,” Samaha notes. Searchable databases and refined decision engines make it possible for CUs to approve auto loans in seconds for complete strangers sitting in dealer showrooms, he illustrates.
“But those are static applications built by humans without any real cognitive function,” Samaha qualifies. “They work. They recognize and respond to preset patterns, but they don’t grow unless a human fixes them, and they don’t solve problems. There’s a valuable level of AI that they don’t reach.” Let’s say current applications of AI have an IQ of 115. The goal has been raised to 160.
Much of the smart automation Suncoast Credit Union has provided to members so far has focused on streamlining mobile and online banking menus and the clicks it takes for a member to get what they want, says Ted Hassenfelt, CIO of the $9.2 billion CU, Tampa, Florida.
Going from five clicks to four used to be a victory. Now Suncoast CU is moving away from menus and clicks as mileposts for improving the member experience, relying more on native features in mobile devices like screen tilt and finger swipes—and eventually to no physical interaction at all, Hassenfelt reports—“anything to remove friction from the member experience.”
AI solutions need to be comprehensive to really make an impact, says Stacey Zuniga, VP/financial services for ENACOMM, Tulsa, Oklahoma, a fintech company that provides self-service solutions and communication technology. With one possible exception—explained in the next paragraph—there is no killer app out there using artificial intelligence to pull crowds of members away from unprepared CUs. “Members have their favorite channels, and they want them all to work well,” Zuniga says. “It’s not one high card that makes a winning hand; it’s a combination.”
Alexa Banking
The one possible exception is Alexa banking. “Amazon Echo and Alexa have spread like wildfire,” Kordeleski notes. Voice assistants make banking “phenomenally easy for members, and the development of 5G processing will make communication hundreds of times faster.
“The [Alexa] technology is already familiar to a lot of credit union members, and they want to use it for more aspects of their digital lives every day,” Kordeleski adds, “which includes banking.” But most can’t. Of the 10,000 or so banks and CUs in the U.S., “maybe 30 have the tools to offer it, and few of them are allowing transactions yet,” he reports. There are compliance issues to consider. The Amazon Developer Services Agreement has its own set of security requirements that financial institutions and any Alexa skill provider must meet, for example—and that’s in addition to complying with federal cybersecurity and privacy regulations.
But members are impatient. Digital assistants like Alexa and Siri have changed consumers’ expectations. “Credit unions and banks are not leading the customer experience,” concedes Ben Morales, CCE, chief technology and operations officer for $3 billion Washington State Employees Credit Union, Olympia.
“Customers are interacting with Amazon, Apple, Google and online merchants in ways they don’t interact with their financial institutions, and they are asking, ‘Why can’t my bank or credit union do that?’ It’s a real challenge, especially for financial institutions working with closed systems. We have to be able to write APIs (application programming interfaces) to [integrate digital assistant services with] our core systems, and then we have to figure out how to do it in a way that provides a satisfying member experience.”
Doing that quickly could be a challenge. In 2008, when app stores started carrying mobile banking apps, Zuniga would ask CU audiences how many had mobile apps available. It wasn’t until 2012 or 2013 before any hands went up, he says.
Conversational banking in some form, not necessarily through Amazon and Alexa, is still inevitable, Zuniga observes, and he thinks it will be a sea change. ENACOMM recently inked a deal with Wescom Resources Group, a credit union service organization based in Pasadena, California, that is wholly owned by $3.5 billion Wescom Credit Union, also based in Pasadena. ENACOMM already has an Alexa banking app that it offers directly and will now also offer to credit unions through Wescom Resources Group.
Conversational banking using digital assistants is coming soon for Suncoast CU members, says Hassenfelt—probably by or before 2020.
“We haven’t rushed into it,” he explains. “We want to be sure the voice biometrics are solid before we move. We are looking at … ways to handle the initial authentication so members would not have to speak their password. It would be like touch ID. … We would register their voice for voice biometrics. That means all subsequent authentication attempts could be handled with the member speaking a phrase and voice biometrics logging them in. We are still working all of this out, however, so nothing is set in stone.”
Suncoast CU is working with BiG to implement the fintech company’s FIVE conversational banking technology. “We expect to start testing it late in the third quarter and to go live by the end of the year,” Hassenfelt says. The service will be available through mobile and online banking platforms and at first will only handle such limited activities as funds transfers and balance inquiries, he reports. “As adoption grows and activity proves reliable, we’ll add functionality.
The CU has long been part of BiG Innovation Club, “so we have access to the source code for this platform should we want to take control of it in the future,” he adds. But given the likely near-term expansion of the platform as AI evolves, the CU will probably not take on the development effort itself.
Washington State Employees CU is taking a cautious approach to AI-driven member communication, Morales reports. “We’re introducing fraud alerts as text messages. We’re doing a proof of concept exercise for loan approvals through conversational AI. We’re looking into chatbots. Whatever we do has to be managed centrally so we don’t lose our deep connection to the member.”
Improving Chatbots
A technology with great potential to deepen member connections is automated member communication. The current state-of-the-art offering, notes Richard Crone, CEO of Crone Consulting LLC, San Carlos, California, is a well programmed chatbot that can respond with real-time texts to member questions or requests. What’s achievable today is a bot that can satisfy the member about 80% of the time, with the more challenging contacts rolling over to a live agent for the other 20%, he explains.
Most chatbots currently are improved by people reprogramming them, but the goal is to have them be AI-driven through machine learning: Once a live agent resolves an issue, the next time that issue comes up, the chatbot would be able to handle it on its own based on the agent’s previous interaction, he says.
But that’s far from what CUs are doing today, Crone points out. “Most don’t even have effective chatbots, but they are starting to show up in the biggest banks,” he notes, “and that shortcoming is causing credit unions to lose members, who more and more are expecting to have questions answered and issues resolved automatically through their mobile banking app.”
Progressive CUs are aware of the problem but stymied by vendors that are not up to speed, Crone says. “The core systems should provide it, but they’re … not investing in improvements. Most CUSOs are focused on collective cost reduction, not collective technology innovation.” The need for efficiency and low unemployment rates are driving some innovation, he notes. “A member services representative can handle about five online chats at a time. A bot can handle many times that.”
And chatbots, sometimes multilingual, can enhance member communication. Where there is a messaging routine, like sending collection notices, robots can efficiently become the messengers, Kordeleski says. “What once required screen scraping can now be done by bots, saving tens of thousands of hours in backroom operations,” he notes.
Bots have successfully invaded some CU contact centers, Samaha notes, and basic questions or requests that occur frequently are being answered automatically by text or email messages and sometimes by automated voice responses that sound like Alexa or Siri, he adds. When the questions or requests are completely predictable, the answers can be automatic. When the questions or requests are unpredictable, the bot gets stumped and a person has to take over.Chatbots are already paying off for progressive Suncoast CU, though it’s a challenge. “We have embedded hard-coded logic that can handle routine questions” in situations that can be scripted, Hassenfelt explains. “AI doesn’t always provide the best user experience today,” he concedes. “There are thousands of ways to ask the same basic question. It’s hard to build an AI application that can recognize and respond to them all, but AI is a fast learner and will get smarter pretty quickly.”
Financial Advising
Among the automated product offerings, personal financial planning has experienced slower uptake than expected, Zuniga reports.
“The use case is evolving,” he says. “It seems that people are less interested in creating long-range plans than they are in regulating their financial behavior. They want to know if they’re overspending or have sufficient savings. They want to know if they’re on track for a wedding or vacation they might be planning. They’re less interested on a daily basis in what they’ll have when they retire, but they’re anxious to avoid debt beyond the student loans they might have outstanding.”
Suncoast CU is cautiously bullish on AI-driven personal financial advice, according to Hassenfelt. “We think that will develop with the build-out of a natural voice channel using digital assistants,” he says. “We’ll probably deploy it when the technology is ready.”
Fintechs are solving the problem of data gaps due to the different financial providers a person might be using, he adds. “Companies like Plaid are making it easy for members to link accounts at different financial institutions [through an app] so we can aggregate the data.” With access to such data, Hassenfelt believes financial advice technologies will be thriving “in three or four years.”
Preparing for expanded use of such AI tools is one reason Suncoast CU has brought in Payrailz, a smart bill-pay service provider based in Glastonbury, Connecticut. The CU went live with the platform in June. Hassenfelt sees Payrailz as a digital payment platform that can help members make sound, proactive finance decisions based on their data, not just a bill-pay vendor. “Our old bill-pay product was used a lot but not really growing, so we went with Payrailz for bill-pay and P2P payments. … We picked them because we like their AI strategy. Their AI potential looked good to us.”
A version of a future member experience may be taking shape in a community bank, nbkc Bank, Kansas City. That bank is working with fintech giant Finastra, based in London, which now includes Malauzai Software, to develop a conversational banking bot that uses Alexa for voice, reports Shuki Licht, Finastra’s chief innovation officer. “We use artificial intelligence and multiple APIs developed by fintechs to go beyond routine banking transactions,” he says. A customer of that bank could soon get natural language answers to questions like “Can I afford to spend $300 on a watch?” or “What would be the approximate monthly payment if I bought a $25,000 car?” “The machine can make recommendations just like a human would,” he concludes.
Nbkc Bank worked with Finastra on the product that was presented in a pre-recorded demonstration at the May 2019 FusionOne conference, according to Eric Garretson, CFO/fintech strategy leader at the $800 million bank. The software did give appropriate natural language answers to personal financial questions. Now nbkc is getting ready to test the product for roll-out to its customers. “We’re eager to partner with fintech companies to see what we can give our customers,” Garretson says. “It’s in our DNA. We’ve been the first customer of two technology companies.”
That’s fantastic. Or maybe not. The benefits of robust AI for financial institutions could be overrated, Zuniga cautions. True artificial intelligence supported by machine learning is not fully utilized for retail banking today, he says, and that’s not necessarily a shortcoming.
“Applications today quantify and clarify, but you don’t necessarily find them changing algorithms automatically. Essentially, they learn and report; they don’t take action independently. Until implementation of AI technology in banking matures quite a bit more, I don’t think credit union managers want the retail banking systems’ ‘brains’ to grow without appropriate curbs. There are a finite set of activities in retail banking that smart computer programming needs to support. And a truly smart solution needs to also observe limits.” cues icon
Richard H. Gamble writes from Grand Junction, Colorado.
