The path to personalization in automation

During the early days of the COVID-19 pandemic, Bank of America’s virtual assistant, Erica, was hammered with requests related to COVID thousands of times per day, but Erica hadn’t been programmed to answer those questions, or even understand their context.

Within one weekend, the $2.2 trillion bank was able to use machine learning to adjust Erica to handle pandemic-related questions, said Hari Gopalkrishnan, client-facing platforms technology executive at Bank of America, at last week’s Bank Automation Ignite event.

“No one asked, ‘Can I apply for a PPP loan?’ They’re saying, ‘My small business is in trouble — how can I get help?’” Gopalkrishnan said. “How do you take that pretty random set of sentences, of words, and distill into what it is what they’re asking for — small business trouble loans, PPP program — and then provide an experience in the back of it?”

Any bank discussion on the “personalization” of the customer experience — how automation can help banks achieve a personal touch when connecting with millions of customers — usually starts with chatbots.

A chatbot typically is introduced to customers through a text interaction and uses decision trees and keywords to respond to them, said Richard Weeks, senior vice president and head of conversational experiences and capabilities at the $553.4 billion U.S. Bank. Weeks joined a panel discussion of new use cases for virtual assistant development and deployment at BA Ignite.

Panelists differentiated chatbots from virtual assistants such as Erica according to their ability to personalize and respond to language; while chatbots offer some personalization and automation, they lack the “wow” factor of virtual assistants, Weeks said. AI-powered virtual assistants, by contrast, leverage natural language processing, which supports more advanced personalization and allows the bot to handle turns in the conversation. Chatbots “can offer some links to FAQs, but it’s not really great for discrete content and the presentment of content,” Weeks said.

Personalization through natural language processing

Banks are increasingly moving beyond simple personalization to more advanced tools that support more automated functions, BA Ignite revealed.

For example, whereas a regular chatbot would stumble if the customer said, “I need to send $150 to … oh wait, actually let’s make that $250.” However, AI-powered virtual assistants can handle that context and respond accordingly, Weeks explained.

This makes the customer interactions “much closer to a conversation,” Weeks said. And while such automations are more complicated to build and maintain, “the breadth and the level of robustness that you can create is much greater within the virtual assistant space.”

That capability means customers don’t have to learn the bank’s menus and applications because they can simply tell a virtual assistant what they want to accomplish, and it understands, Gopalkrishnan said.

The natural language processing used by AI also allows the virtual experience to handle more complex situations automatically, without the need for human intervention, as Bank of America did during the pandemic with Erica.

And customers appreciate the personalized experience offered. Bank of America’s weekly customer satisfaction surveys show that 90% of customers enjoy the experience Erica provides, Gopalkrishnan said.

“Our client can talk to us in any way they want to,” Gopalkrishnan said. “There are a thousand ways they could ask us to pay a bill. They don’t need to learn our language — we learned theirs.”

Personalization as experience

Personalization doesn’t have to happen only with chatbot-customer encounters, however.

Justin Hunsaker leads the omnichannel experience at KeyBank, where he heads the originations and wellness team in enterprise digital. At BA Ignite, he explained how the $170.3 billion bank is transforming customer interactions to combine the best of the old and new worlds to create what he calls “phy-gital experiences.”

Millennials are starting to sign mortgages, Hunsaker pointed out, and KeyBank’s research found that this generation is most comfortable with a combined digital and physical experience.

“It’s not about just telling somebody the answer,” Hunsaker said. “It’s about giving them access to the information, and then also telling them your recommendation. And when you put those things together, millennials seem very comfortable with that.”

The Cleveland, Ohio-based bank provides that combined experience through its new in-bank system, KeyBank’s Financial Wellness Review, which leads the customer through a series of questions alongside a banker, then makes recommendations for the banker and customer to discuss. Customers who interacted with bankers this way gave the bank higher net promoter scores, which measure how likely customers are to recommend the bank to a friend, Hunsaker added.

Business and commercial bankers, on the other hand, don’t want digital tools to interfere with the expertise they provide to their customers, Hunsaker pointed out. For those cases, there are other places to automate the user experience, he said, so that the personalization and automation fit the situation. It is affordable for business bankers to work closely with key business clients, so there is no real need for digitization or standardization, in that case, Hunsaker said. “But there is a huge need to make things simple, fast, easy, and we’re doing a lot of work in that space as well.”

Personalization as product

For Cross River Bank, the personalization isn’t just built into the product — it is the product itself.

The Fort Lee, N.J.-based bank primarily works with fintechs and midsized enterprises like Affirm, Rocket Loans and Upstart, to underwrite more than $1 billion in loans per month through apps while serving a wide range of borrowers. Cross River, founded in 2008, has raised a total of $361 million in funding, according to Crunchbase.

“We’ve figured out how to optimize technology to figure out to say ‘yes’ to a borrower that fits a kind of defined box, and do so at scale and automate the funding and the underwriting and everything else that goes along with that,” said Jesse Honigberg, senior vice president and technology chief of staff, at a BA Ignite session on personalization at scale.

“That was really a collaboration between our front office or back-office client service teams, to figure out how we can understand what good enough is, and how we could try to serve the most borrowers by not trying to serve every borrower,” Honigberg added.

Similarly, the goal of financial technology and bank holding company Green Dot is to target low- to moderate-income customers — many of whom are unbanked,” said Abhijit Chaudhary, general manager of Green Dot’s direct-to-consumer business. Many of Green Dot’s customers have $20-$40 in their bank account, if they even have bank accounts, he added.

“How do we create a banking product that really helps [customers] automate their money movement?” Chaudhary said, adding that it has to be as simple for users as “going to a retail store, like a Walgreens, Walmart, just picking up a card at the shelf and loading cash right then and there.”

To that end, the fintech’s new mobile app product, Go2bank, is personalized to meet the needs of unbanked customers, and offers up to $200 in overdraft protection, plus high-interest savings, and a reward system. The fintech, which launched the product in January, spent months learning what its customers wanted, which was fast access to their money, Chaudhary said.

“When you talk about automation, and how it helps this audience, it’s important to understand what they care about,” Chaudhary said. “They care about having access to their funds immediately.”

Green Dot also provided funding of almost $3 billion to millions of its customers during the third stimulus round, Chaudhary said. “We did this five days before any of the traditional banks did.”