Virtual assistants have been primarily the domain of large banks until now, thanks in part to offerings made through core providers.

BankSouth, a $1.2 community bank based in Georgia, began leading the way in November 2021 with its adoption of a chatbot named RITA — or real-time interactive teller assistant — driven by conversational artificial intelligence (AI). The bank implemented the chatbot as a way to handle after-hours calls and take some pressure off its call center.

RITA is a “fantastic example” of how conversational AI can be used at smaller banks, said Lindsay Soergel, chief product and customer experience officer at Kasisto, the chatbot’s developer.

Kasisto builds AI virtual assistants for financial institutions (FI), historically working with larger banks such as TD Bank, Westpac and JPMorgan Chase, Soergel said. But that is starting to change, particularly as banks can now access Kasisto through core provider NCR’s digital banking platform, she noted.

Financial services fintech FIS will also soon go live with the same technology through its digital banking platform, Soergel said. Currently, Kasisto has about 15-18 projects in various stages of deployment, she added.

“This year is the first year that conversational AI and chatbots have become one of the top five spend categories,” Soergel said. “Even still, just because the folks have decided to start looking at spending money, and it doesn’t mean a lot of folks have necessarily brought things to life, so it is still nascent.”

Embracing chatbot technology

BankSouth has embraced RITA not just as a technology tool, but as the digital face of the bank. RITA is included in advertisements and her cutouts can be found at its five physical branches. She even has her own email account, which is used to respond to members who interact with her, said BankSouth Chief Innovation Officer Harry Gunsallus.

“We needed to make sure that something called a ‘conversational AI’ — which could be off-putting to some folks — had to be approachable,” Gunsallus said. “So we thought: How do we make a chatbot more humanistic? Somebody that you could actually have an interaction with, somebody that you could have an emotional connection with.”

Employees participated in a chatbot-naming contest before a persona was created. The bank wanted an omnipresent chatbot that would be fun, lighthearted and a little sassy — like a favorite aunt, Gunsallus said.

“She’s always positive, interested, smart and savvy with a dash of sass,” he said. For example, if asked for her address, RITA quips, “Isn’t it a little early in our relationship to be sharing addresses?”

RITA’s foundation

Kasisto’s AI underpinnings originated at the Stanford
Research Institute, the home of Apple’s virtual assistant Siri. The vendor moved AI capabilities and models to a multi-tenant cloud when the technology was made available through core provider NCR. BankSouth’s RITA resides within Kasisto’s platform with “a couple of dozen NCR clients,” Soergel said.

That allows the vendor to achieve economies of scale with the technology, Stephen Epstein, chief marketing officer at Kasisto (Editor’s Note: Epstein has since left Kasisto), previously told Bank Automation News.

“Price has been a barrier to entry for a lot of the small financial institutions to get access to this robust technology, and that’s no longer there,” Epstein said.

The AI is trained with out-of-the-box intents, or functions, designed specifically for FIs. Creating RITA was as simple as turning on a checkbox and deciding which trained questions and responses to use, Gunsallus said. There were some intents that weren’t needed by BankSouth, such as those related to stock options and trading, which the bank does not offer. The only intent the bank needed that was not available was related to tax notices, which Kasisto then built with input from BankSouth.

“Out-of-the box it came … 500 Q&A intents in responses. There was a couple of weeks of training,” Gunsallus said, adding that the process took about four hours a day for six weeks. “We had myself and one other person, and we would bring in subject matter experts from the areas to refine the responses.”

Thank you, RITA

BankSouth turned RITA on live for employees first, who hammered it with questions, Gunsallus said. Since then, public responses have been surprising.

“In the chats, we’ve actually had people thank RITA, which I think is hilarious,” Gunsallus said. “They thank her for her answer because they believe that there’s a real person behind those answers.”

While RITA was initially intended to answer after-hours call center queries, BankSouth has also built an account onboarding process onto the RITA platform that advises new customers on how to complete their account setup. Previously, customers would start the process without the required documents, which would result in their login timing out. Since RITA now offers clues about preparation, the bank’s abandonment rates for that automation have disappeared, Gunsallus said.

“We’re getting lifts, not just from a call-center perspective, but we’re getting lifts in terms of customer interaction and engagement,” he said. “We see people actually engage more with the personification than they would just a blank email campaign.”

RITA is also working out well in her intended task of fielding customer inquiries, Gunsallus noted. Out of 16,000 requests handled online, only 309 escalated to a call center employee, an average of 4% of calls each week.

“RITA is crushing it,” he said.