Are You My Banker, or a Computer?

Editor at Large

USAA is using virtual agents, personalization and artificial intelligence to forge what it calls a "digital empathy" with its members – conversations that feel human and personal, with an understanding of each customer's behavior, likes and dislikes built in.

If empathy – the feeling that you understand and share another person's experiences and emotions – sounds like a tall order for a financial services company that almost never interacts with customers in person, it is. But it's a worthwhile goal for USAA and would be a smart objective for any financial institution.

"Most banks, especially the large ones, need serious help in being perceived as being more empathetic," said Ron Shevlin, director of research at Cornerstone Advisors. "That's not news to them; they understand that. USAA is absolutely on the right track for doing this."

It makes sense for USAA to lead the way here. As a direct bank, it has few branches – it will soon be down to four. Human interaction generally takes place through its call center.

"The No. 1 thing our members say about our call center when they call in is, 'The member service rep knew me, they empathized with me.' When you hear words like that, you think, how do I take that and create digital empathy?" Patrick Kelly, assistant vice president of emerging technologies at USAA, said in an interview at the Banking Analytics Symposium last week. "If you have a friend you've known for a long time and they know things about you that the general public doesn't know, you have a tendency to be loyal to that individual. You trust them more; you'll keep them around in your life for a long time. The same is true of your institution."

Shevlin recalled visiting USAA a couple of years ago and listening to an agent's call with the wife of a deployed military member.

"It was an amazing phone call," said Shevlin, who is not known for fawning appreciation of anything. "The call was close to an hour." The agent patiently worked through the customer's complex problems. It's important for USAA to be able to offer a similar experience to customers through its digital channels, he said, to provide consistency.

USAA's track record suggests its goal is achievable.

"Lump all the top U.S. retail banks together, and USAA will often be found at the top in measures such as adoption, first-to-market and design," said Jim Van Dyke, founder and CEO of San Francisco consultancy Futurion.

"Speak with their various digital leaders and you see internal traits that similarly set them apart: intense focus on digital leadership, empathy for the customer or member, and a passion for … great customer experience," said Van Dyke, who was also one of the original founders of Javelin Strategy & Research. "It's really that dramatic of a difference; at USAA the members of the provider team simply devote more time and passion to talking about the people they serve."

Building Blocks

One component of empathy for USAA is the ability to communicate in plain language over digital channels.

"Banks historically don't communicate in language that anybody else uses," Kelly said. "We have our own vocabulary. When you can communicate to somebody in the tone they like to be communicated in, it can go a long way."

Another piece is personalization. USAA is teaching its machine-learning systems by feeding them small data sets about individual people. This is almost the opposite of IBM's Watson, which ingests massive quantities of data and documents to learn.

"If you want to create a one-to-one experience, that data set is very small," Kelly said. "The cohorts you think of – the people who are like you at that level of personalization, are very small. So the data sets associated with it are small. That's the struggle side of AI now." The less data a machine-learning system has to learn from, the longer it will take to get it right.

An important facet of personalization is learning when to communicate with customers, Kelly said. When a person is standing in line at Starbucks or driving to a doctor's appointment, they're unlikely to want to hear or read an offer for life insurance.

Humans have an ability to pick up clues that is tough to instill. "If you're on the phone with somebody and you hear a baby crying in the background, you can tilt your conversation accordingly," Kelly noted. "A computer can't do that. If I know you never bought a white car in your life, why would I ever show you a white car when you're searching for cars?" USAA would like to teach its artificial intelligence software to learn these subtleties.

This requires an understanding of data that is rare today. USAA plans to track customers' behavior, store it and apply that knowledge to offer a piece of advice or best action the next time the customer logs in.

USAA's personalization efforts began by letting customers drag and drop the menu items they prefer to see on their mobile apps.

At the next level, USAA is automatically manipulating mobile device and website views based on what the system knows about the customer. For instance, if a member has filed an insurance claim, the claim status might appear on the top of the page.

Someday USAA may fashion an individual product for each of its 11.7 million members. "When you do that, you'll have a customer or a member for life," Kelly said.

A third, intertwined thread of digital empathy is USAA's virtual assistant, which Neff Hudson, who is now vice president of corporate development, launched two years ago.

"It's getting more accurate every day," Kelly said. "You get the data from that, and the data helps you craft the next product, the next use case for it. And people get accustomed to using it, so then the product developers look at it and say, what can I do with this that I hadn't thought I could do with it?"

When members speak to USAA's virtual assistant, a natural-language-processing engine listens to their questions and translates them to text. In the beginning, the virtual assistant sent all the questions to a team of people who composed answers that were logged in libraries.

Now, when a customer poses a question, the virtual assistant searches the answer libraries. If an answer exists, it returns that. If the customer has requested a simple action, such as block or unblock a card, the virtual assistant can do it. If the assistant can't find an answer in the libraries, it will route the customer to the help desk.

In the early days of the virtual assistant, people often asked questions like, I've had a fender bender and I'm on the side of the road, what should I do?

"We still have a lot of that," Kelly said.

'Conversational Commerce'

Kelly would like to have the virtual assistant become a vehicle for "conversational commerce," a term that is widely attributed to Uber developer Chris Messina, who wrote a blog about it for Medium, but was originally coined in 2009 by Dan Miller, cofounder at Opus Research.

Miller defines it as "a combination of technologies that facilitate person-to-person, person-to-machine and machine-to-machine conversations to promote commerce." They take place over the phone, through chat, across messaging platforms and the like.

Kelly sees this as part of a broader trend in the evolution of apps themselves.

"In 2014 was the era we call the 'appsplosion,'" Kelly said. "Everybody had hundreds of apps on their phone, every app did one thing. That was a very noisy solution for a lot of people; they had to remember what app they needed, the login credentials for that app."

Attention spans have shrunk; you now have eight seconds to capture a millennial's interest, he said. "People have begun deleting apps from their phones at a higher rate than they ever have," he said. The trend has moved beyond single-use apps; people want "solution" apps.

"Google gives you access to the entire internet through one search box," Kelly noted. "They don't give it to you through a menu structure, the way most financial services companies do. With natural language understanding and cognitive computing AI, we're moving to where we can give you access to the entire USAA services through one search engine."

For any company that's not Google, such multipurpose "solution apps" run the danger of becoming too large and unwieldy.

"You have to be ruthless with what you're willing to build native and what you want web-based," Kelly said. About 70% of USAA's app is still web content delivered to a mobile web page, which allows the app to be lightweight. "Now you have to decide, what experiences are so important to your brand identity and your users, that you want to make them native?" In USAA's case, mobile deposit is hard-coded into the app; many other actions occur on the mobile web.

The Trouble with Bot-Speak

One challenge for USAA and others is, while technology is pretty good at handling a written conversation, automated voice conversations are much harder for software to handle and rarely seem natural.

"When you're having a text-based conversation with an AI agent, you often don't know you're interacting with tech and not a live human being," Shevlin said. "But on the voice side, you still feel like you're interacting with Siri or Alexa. Maybe there are some folks who would rather interact with Alexa and Siri than real people. But I'd think there are still plenty who are put off by that."

Virtual assistants can handle simple requests, such as account balances, with ease. But they can stumble on complex questions involving, say, financial goals. It's the more complicated questions that bring up the need for empathy, of course.

Overall, Kelly thinks it will take two to five years for artificial intelligence to be advanced enough to do all the things USAA wants it to do. But because of AI's ability to continuously learn, it will be significantly ahead of many competitors when it does get there.

Editor at Large Penny Crosman welcomes feedback at penny.crosman@sourcemedia.com.

For reprint and licensing requests for this article, click here.
Bank technology Mobile banking Digital banking Analytics Consumer banking
MORE FROM AMERICAN BANKER