Teaching Chatbots To Speak The Language Of Consumers

Capital One Eno

Although Eno is a new way for Cap One and its customers to communicate, Ken Dodelin, VP of digital product development at Capital One, says that it’s really not all that new. He tells Karen Webster that it simply builds on the way in which they’ve always communicated via text. So, how does one teach a bot to speak emoji? Dodelin says, it’s all about the context.

Yes, technology is enabling machines to communicate more efficiently with humans.

But truth be told, that communication isn’t always smooth.

While artificial intelligence interfaces like Amazon’s Alexa have come a long way, there are way too many times where consumers still feel like they’re talking to a robot that delivers a lot more misses than hits.

But in the creation of its own chatbot, Capital One decided to take an entirely different approach. The company recently announced the launch of its new text message-based assistant named Eno, designed to help customers manage their credit card, checking and savings accounts.

Ken Dodelin, VP of digital product development at Capital One, joined Karen Webster to share what makes Eno unique and why meeting customers in the mobile text messaging channel was so critical for communicating in an effective way — since it also happened to be the way that consumers were communicating with the bank in the first place.

Embarking on Natural Language

Eno, considered to be the first natural language SMS chatbot from a U.S. bank, is a gender neutral assistant-like platform that enables Capital One customers to access information about their financial accounts through a simple text conversation.

Though Eno is a new feature, Dodelin said that its development is a variation on a theme that was at the heart of how consumers and Capital One were already interacting.

“The funny thing is, we actually launched the chatbot long ago, but we just didn’t know it,” he explained. “What I mean by that is, we had these two-way SMS alerts where we would let customers know if there’s something important going on with their account, or we might ask them to send a text back to confirm or deny.”

At the time, that SMS service only understood a very limited number of responses from customers, but Dodelin said customers still sent back all sorts of messages, including explanations, questions and even emojis.

“Eno is our first step toward engaging with customers in these conversations that they’re already having with us,” he added.

Finding Value in SMS

In today’s world, if a company wants to be able to continually communicate with its customers, Dodelin said that it has to not only speak their language in the way they want to talk, but also be accessible within the channel customers want to chat in.

Since texting is already a natural communication mechanism for so many consumers, Capital One is now creating a natural language interface that allows those customers to do more than just respond to an account alert.

While Eno is able to handle providing customers with information, such as their account balances, recent transactions and accounts summary within the confines of an SMS conversation, there are certain instances where it will provide the customer with a link to Capital One’s mobile site in order to complete an action.

As Dodelin pointed out, Eno should be viewed as complementary to Capital One’s existing mobile app, enabling customers who value speed of access to quickly have their questions answered within a channel they are typically already using throughout their day.

To enable Eno, customers are required to login to Capital One’s website and grant permission for these conversations to happen. With the exception of being able to pay a bill, the use cases enabled for Eno are informational and are also the same use cases currently available through the SMS alerts, just now in a “much more accessible and user-friendly natural language interface,” Dodelin added.

He noted that if the day comes where Capital One decides to enable higher risk transactions within Eno, then it would also include step-up authentication in order to ensure that data remains protected.

The Chatbot Learning Curve

Natural language in SMS is very different that natural language through a chatbot like Alexa or Cortana. In text messages, people’s natural behavior is to use emojis and shortened phrases, which Dodelin said is an opportunity for Capital One to shift the paradigm of bank speak.

“Now we are designing to be able to understand how customers interact naturally in this channel and design banking conversations around that, versus trying to move them into bank speak,” he said.

Part of that design process also involves ensuring that Eno has emotional intelligence and understands the context of a conversation. If somebody uses a thumbs up, it’s imperative that Capital One knows if they’re referring to that particular interaction, or if they’re referring to their feelings about Eno, Dodelin noted.

In trying to understand what kinds of things customers want from Capital One and the way that they’re talking about it, the bank can use that data to inform what features or capabilities it might design into Eno in the future.

Dodelin said there will be a great deal of growth in Eno’s understanding as time goes on and that Capital One views dialogue management as an area where it will continue to build on Eno’s capabilities.

From the use case perspective, Capital One will be able to utilize the interactions Eno has with consumers and knowledge it gains in order to continually develop its own product roadmap.

Looking into what’s next for Eno, Dodelin said the platform could certainly borrow from SMS and utilize it in other channels as well.

“That’s an exciting opportunity — to go where the people are and not make them all come to us in one place,” he added.