Far from changing banking, chatbots aren’t keeping up with call volumes in the COVID-19 era

Before the coronavirus pandemic, online small business loan provider Kabbage got a few thousand phone calls a day from customers and potential customers. Now the Atlanta-based fintech is receiving hundreds of thousands of calls a day, many of them related to government-backed small business loans under the Paycheck Protection Program. At times it feels overwhelming, said its cofounder Kathryn Petralia.

“It would have been impossible to manage without automation,” she said, explaining that the majority of calls go to an interactive voice response system that can disperse some information and, ideally, direct callers to the appropriate people and departments.

Kabbage is not alone in trying to shift more of the burden of customer service from humans to machines; nor is this desire new. In 2019, chatbots saved banks an estimated $209 million in 2019, according to Juniper Research. But now, the pandemic, which has sharply increased call volumes across the financial sector, is putting the growing role of artificial intelligence in customer service to the test, as more callers are forced to rely on chatbots and digital assistants, or wait on hold for hours.

Growth and adaptation of technology that would have normally taken years has happened in just a few months as banks struggle to deal with increased volumes of queries, and quickly realized their human staffs couldn’t meet the need.

“COVID made it clear for banks that human-powered call centers are not the only answer,” said Zor Gorelov, co-founder and CEO at Kasisto, which makes KAI, a customizable and AI-powered customer service solution for banks and other financial companies. “Banks now understand that virtual assistants are no longer a ‘nice to have’, but a ‘must have’.”  In addition to increased call, web and app inquiries from clients and customers, banks also faced labor shortages in their call centers as many countries around the world issued stay-at-home orders that did not allow employees to report to work, he said. In some places, many call-centers employees fell sick with the virus, Gorelov added.

TD Bank was one of many institutions that turned to Kasisto and others to quickly deploy AI-based solutions to deal with volumes, and added a chatbot to its app in a matter of weeks, Gorelov said.

Kasisto says its platform can answer 85% of queries, and has allowed some clients to cut call center staff by 50%.

But there is still a long way to go, with many institutions displaying messages about long phone wait times on their homepages. Despite the excitement around virtual voice assistants, like Erica at Bank of America, Eno at Capital One, they are not completely replacing humans — and probably never will, said Daniel Faggella, CEO of Emerj Artificial Intelligence Research. 

“At this point, the science of natural language processing is not up to the level that we can call actual conversation,” Faggella said  “At its best, the biggest value AI offers now, if you want to know anything more complicated than your account balance, is the ability to route calls to the right place.” A good virtual assistant or chatbot clarifies what someone needs and transfers them –and the relevant background information — to a human.

Faggella says that the technology will inevitably improve, but this depends on data scientists working closely with customer service experts to properly develop and effectively deploy the solutions.

“There is plug-and-play AI that is going to help with the phone call situation,” he said.  He also added that people will always be part of the picture of customer service, especially for high-value clients.

But customer-facing AI can have additional benefits. Gorelov said that AI is helping institutions quickly learn what people want. By tracking data about queries, banks can see what customers want to know, and can quickly train AI systems and update less-advanced channels like online lists of frequently asked questions to make sure they stay relevant. 

“All of a sudden, in a matter of weeks, we went from questions like ‘What is COVID-19?’ to ‘How often do you clean your ATM machines?’’ he said. Relying on individual people to report to management about which questions are new or most popular among customers would have taken much longer, he said. He also added that some customers are more willing to divulge information, like admitting to an addiction or debt, to a chatbot or virtual assistant than to another person.

Customer-facing AI is also showing up in other applications, including Intuit’s recent launch of Intuit Aid Assist, a platform that helps small businesses navigate the options for federal assistance offered by the Coronavirus Aid Relief and Economic Security Act. 

“Intuit Aid Assist codifies the complexity of the hundreds of pages of the CARES Act into an easy-to-answer user experience with five to ten questions,” that let users know what aid they are qualified for and how to apply, an Intuit spokeswoman said. 

But despite these developments and the media spotlight on banking chatbots and virtual assistants, Faggella emphasizes that cyber security remains the number one field for banks’ investments in AI, and offers the largest returns on that investment.