By John San Filippo
Finopotamus was onsite at Money20/20 USA in Las Vegas Oct. 24-27. Co-founder John San Filippo interviewed more than two dozen technology experts on a wide range of topics. The results of these interviews are presented in this series called “The Voices of Money20/20.”
Jake Tyler, co-founder and CEO of Finn.AI, took time out from his busy conference schedule to discuss the benefits of artificial intelligence (AI)-driven chatbots with Finopotamus. He started the conversation by first defining the problem that Finn.AI seeks to address.
“It used to be that if you wanted to get something done at a bank or credit union, you could go in and talk to somebody and get it done,” he told Finopotamus. “But today, most of our banking is getting done on digital channels. We recently did a survey of a thousand U.S. consumers with MX and a few other companies and found that over 80% of millennials and Gen Zers rely on mobile as their primary channel.” He said that the challenge is getting the same things done on “a tiny little screen” that people used to do in branches.
“The benefit of a conversational interface is that you can get things done just by asking inside your digital channels,” said Tyler.
Putting AI to Work
Asked about the role of AI in Finn’s product, Tyler answered, “Everyday human conversation has two parts. One, do I understand what you're saying? And two, do I have an answer that makes sense in the context of that conversation? AI is applied in our system on the understanding side.” He added that Finn’s AI engine was built and “trained” on millions of real-world banking conversations over the past five to six years. This, he said, means that “we're able to understand most banking conversations out of the box today.”
He added that the part that's not AI, but is still complicated, is having the right answer. “Asking for a balance is a pretty simple question for people to understand, but getting the answer can be a little bit more complicated,” said Tyler. “You have to either point someone to where they can find that balance or make an API call to retrieve that balance. API connections make bots really powerful.”
According to Tyler, financial services present a unique challenge. “The challenge is that in such a highly regulated space, you don't want the AI to do something wrong,” he noted. “Our answers live in a content management system on our side. Those answers are reviewed by our compliance team, as well as by the credit union, so we know exactly how the AI is going to respond to any given question.”
Better Than Ever
Asked to compare AI-driven systems to legacy keyword-driven systems, Tyler said there isn’t a comparison. “Think about trying to use voice transcription on your phone 10 years ago,” he said. “It was terrible. It wouldn't understand you. It was basically useless. Now I use it all the time. It's great.” He added that because of the tremendous advancements on the understanding side, the user experience has been improved by magnitudes.
The challenge for Finn.AI at the outset was amassing enough data to create a usable system. “We started with essentially no data,” said Tyler. “We’d make up different ways of how people could ask about their balance, for example. That got us to market and now we have a system that continues to grow and learn. Now our AI is trained across the aggregate of data of all our customers. This means we have a very big dataset.”
As for results, Tyler claimed that credit unions deploying the Finn.AI system can see measurable improvements in just the first month after deployment. “The results that we typically see in month one are somewhere between 60 to 80% of conversations resolved without a human (getting involved).”
He said that in talking recently with one of Finn.AI’s credit union customers, that credit union’s “call volume went down, but length of calls went up, which is what we would want to see because our system is handling the repetitive, routine stuff. It's the basic stuff that bots are good at.”
Finopotamus then asked about integration with existing credit union systems.
“There are three integration points,” said Tyler. “First, we're normally integrated with the contact center in some way. AI lives there to handle your level 1-type inquiries – the simple things. The second integration point is the digital banking channel, not necessarily integrated to APIs at this stage, but living in the channel via our widget. Then over time, what we see with our customers is they all add some API integration, normally to the digital banking layer, not directly to the core, although it depends on the FI. That’s the third integration point.”
Tyler explained that Finn.AI has partnered with Q2, among other digital banking providers, and pre-integrated with some of its APIs and digital channel. However, at one of its larger credit union clients, Finn.AI integrated its system to the credit union’s middleware layer. He added that the middleware platforms it currently integrates to include ModusBox and NXTsoft. He also mentioned Tyfone as an integration partner for digital banking.
“If you're a bank or a credit union looking at a product like this today, what I can tell you is, it works,” concluded Tyler. “It's been an evolution over the last few years, but today you can expect something to work and deliver value quickly. The trick is to partner with a company that has the understanding side already done, because that’s most of work. We know because we did it.”