Eno: Designing a Financial AI That Recognizes and Responds to Emotion

Rewind one year. Imagine a team of people sitting in a room together and pondering a question: Could customers ever trust a chatbot enough to talk with it about their money?

Six months later, that same team takes a stab at answering their anthropological question by launching Eno — the first natural language SMS chatbot by a U.S. bank — at SXSW in Austin.

Eno is a gender-neutral bot who understands emojis, has a backstory guiding its character quirks, and is willing to drop the occasional pun. Throughout the pilot over the last six months, the Eno team learned a couple compelling things that reinforce the power of an “AI with EQ” — a phrase coined by Scott Karp (Director of Conversational AI Products) in a meeting one day — unique to the conversational interface.

The Emotion in Human-Computer Interaction

For years, we’ve reflected gratitude in our button nomenclature within the GUI. Whether you’re at an ATM or in a mobile interface, you may see a “Thank you” or a “No, thanks” written in a button label. That’s a conscious decision the writer-designer is making to insert politeness into the interaction. It can be delightful in some contexts and grating in others (particularly if you’re frustrated), but the intention is generally genuine.

So when we launched Eno, we weren’t sure how customers would respond once Eno was done handling their request. Would they just stop talking and walk away? What if Eno didn’t have the answer they needed? We aren’t forcing them to respond, after all… it’s an open-ended conversation in a natural language interface.

Surprisingly (and to our team’s delight), within just the first 10,000 conversations Eno had with customers, “thank you” was in the top 3 most-texted phrases. 

That comes in many forms, from “thanks” or “tx” to — in more extreme examples — heart emojis, confetti, and even marriage proposals.

screenshot of white search bar with multiple suggestions listed in bold black text underneath

Customers seem to express gratitude to Eno no matter the use case, too — it’s not just when Eno substantively provides information, like providing the balance in a quick exchange or making a credit card payment. People say thank you when Eno responds to personal questions or uses one of those aforementioned puns. It’s not lost on the team when a “You brightened my day” comes through; that’s a real person, unprompted, being polite to Eno. Expressing gratitude.

This pattern of thanking Eno, or saying goodbye to Eno, or telling Eno “I love you” — no matter the use case, length of conversation, or number of conversations over time — is a signal we’re achieving something greater than just being factually correct (although that’s absolutely crucial). We’re also striking a chord with Eno’s warmth, humility, and relate-ability as a character representing our company’s values.

screenshot of white search bar with multiple suggestions listed in bold black text underneath

That’s positive momentum and learning for our team, who’s committed to continuing to build out Eno’s emotional intelligence (EQ) as we add more conversations to the mix. Maintaining that emotional connection is paramount.

Customers are banking with emojis

Customers can get a quick report of their accounts by texting Eno any form of “balance.” They also can get the same by texting Eno a money bag or $ emoji, or the plain-text $.

Turns out, SMS behavior between a customer and their bank isn’t much different than between a customer and their social circles; they use emojis about as often as they use the word “balance.” What this teaches our team is that customers aren’t just learning new interaction patterns for banking, but applying the ones they already use.

Emojis convey short-hand SMS language (money bag) and emotion (heart or wave), and that means we’re learning new ways of communicating that already are tailored to people’s behavioral patterns for that channel.

And so that opens up an entirely new world of possibilities to explore, as evidenced by others who similarly see the power of emojis as a natural short-hand to prompt new UX interactions conversationally.

So with Eno now being open for business to millions of Capital One customers, the team behind the experience is thrilled to be continuing the journey of reimagining how customers communicate with Capital One about their money; to keep pushing the boundaries of communications that are as natural, accessible, and frictionless as possible — and all powered by Eno, an AI with EQ.

Learn more about Eno, and sign up here.

Steph Hay, Vice President & Head of Conversation Design, Capital One

Product design leader versed in AI and machine learning. Pioneered content-first design and Lean Content testing, two low-risk methods for proving traction before building a product. Co-founded FastCustomer and Work Design Magazine. Made 1nicething.com.

Related Content