AI And Credit Unions: A Tool, Not Just Tech

Across the U.S., there are 5,757 credit unions with 103.992 million members – and it’s been reported by the Credit Union National Association (CUNA) that roughly 45.4 percent of the nation’s financially active adult population receives some financial services from a credit union.

There’s a reason for that, Fotios Konstantinidis, SVP of fraud products at CO-OP Financial Services, told PYMNTS in a recent conversation. The portfolio of financial services offered by those credit unions often comes with a fairly distinct flavor. Consumers who prefer banking with a credit union, he noted, are there for products that meet their financial service needs, but they are also seeking a provider with a reputation for fostering closer, more personal relationships with their customers, who they call members.

Those members seem quite pleased with that relationship-centric approach to banking. According to the latest edition of the PYMNTS Credit Union Tracker, 81 percent of credit union customers report being highly satisfied with services received.

Some of that satisfaction is related to moves made by players like CO-OP, pioneering the shared branching concept that enables the member of one credit union to use ATMs and other banking services at any other credit unions in the United States that are also part of the network. And then there are the efforts made to develop better online and multichannel offerings for their increasingly digital customer bases.

It’s why, as Konstantinidis pointed out, artificial intelligence (AI) is emerging so quickly as an important new element in the digital lives of credit unions – because its capabilities serve as a powerful tool for developing even closer and more personal relationships with customers.

“AI has become something of a buzzword recently, but it makes sense to start with how the AI tool can be used to best solve the actual problems of customers, instead of with admiring how novel the technology is,” Konstantinidis said, adding that working backwards from the members’ problems that need solving is the only way to align the AI technology with those problems.

Understanding the AI Journey

In today’s credit unions, the use of AI is a mixed bag.

Some (usually larger) credit unions are on the cutting edge, while others are on the launchpad and pretty far into strategic discussions. Among the rest – and probably still the majority, according to Konstantinidis – there is an understanding that working with and building out AI functions is mission-critical, though they are still working through how to make a business-wide strategic move.

“I think people forget that using learning of neural networks as a fraud-fighting tool has been common in the industry for a few years now,” he said. “What is really new is that with the emergence of the cloud, there is a lot more computational power that can scan terabytes of data in a millisecond and render judgements on it.”

While consumer technology is advancing, Konstantinidis noted, fraud is ever-advancing, too – and credit unions are well aware that at this point, staying ahead of fraudsters requires a smart AI approach, as traditional methods are simply no longer sufficient.

But as credit unions begin to build those smarter, more adaptive profiles of their customers to fight fraud – and start seeing the bigger picture of customers’ transaction profiles, instead of viewing each transaction as a one-off event – something starts to happen. That credit union actually starts taking a much more personalized look at their customers.

“You start to see that this member travels a lot, so you don’t flag them for spending in a different zip code immediately,” Konstantinidis shared as an example. “Or you see that another customer is a big spender, or just got married. There are now a lot of marketing opportunities coming into play, to offer them the right products. This is where AI comes into play – it is about the ability to enrich your data [in order to] understand your members in a better way.”

However, he noted, it’s also important that the customer is an active participant in this process. Konstantinidis pointed out that the journey to using AI to enrich customer relationships has to be a mutual one – not one where the customers feels as though they and their data have been kidnapped along for the ride.

Customers shouldn’t be bewildered by offerings, or wondering why their credit union seems to be tracking them so closely. In fact, Konstantinidis noted, transparency is actually a good rule. It’s important to make it easy for customers to know how and why their data is being collected, and to give them the ability to opt out.

“In relationship banking, it doesn’t really pay to make the customer suspicious of you,” Konstantinidis warned.

The Path Forward

It is a good time in history to begin expanding the use of AI and machine learning for credit unions, Konstantinidis noted, because as the cloud has made the processing power necessary for it to really thrive, AI itself has become something of a commodity that credit unions can now work with firms to incorporate, without having to become master technicians themselves.

And now for the “but.”

None of this is nearly as “lightswitch-flip easy” as the buzz around it might lead one to believe. Automation is a miraculous thing, asserted Konstantinidis, but it still requires an awful lot of tedious, manual work.

Konstantinidis said that gathering, sorting and cleaning the data is an essential but tedious first step to obtaining data in a usable form. Often, he noted, AI is portrayed as a big funnel into which data is poured, magically coming out the other end in an amazing state.

“In fact, you spend a lot of time putting data into a format that can work with your AI engine,” he explained.

And that’s just the beginning of the work.

In the early phases, that engine will be fed a lot of data – and will probably fail a lot, which Konstantinidis noted is exactly what is supposed to do, at least at first. The road to a product launch involves feeding AI a lot of data, and training it to “think properly about outcomes.”

“AI learning isn’t magic – it is a human training an AI to perform a function with data overtime,” Konstantinidis emphasized.

Ultimately, he said, it is a worthwhile effort, and one that will likely change the face of credit union banking for customers for the better. And it’s a move that credit unions know they have to make, even if they are still working out the specifics.

According to Konstantinidis, very few of today’s credit unions are currently ahead of the game, but most really understand that this is what they need to boost their bottom lines.

“Credit unions want to understand their members in a better way,” he summed up. “Not only does this help reduce fraud, but it also opens up new revenue opportunities: a double win.”