How A Former Pixar Designer Gave Life To Capital One’s Chatbot

Bot is gender-neutral with a sense of humor and won’t stand for rudeness

When Capital One Financial Corp. was developing its artificially intelligent chatbot during the past year, it drew on the experience of Audra Koklys Plummer, who spent six years helping to create fictional characters at The Walt Disney Co.’s Pixar Animation Studios.

Ms. Koklys Plummer, head of AI design at Capital One, has tried to imbue the bank’s text-based digital assistant, Eno, with compelling character traits like empathy and humor, similar to the way her teams at Pixar developed characters in films such as “Ratatouille.”

.. It also has traits such as a sense of humor, trustworthiness and empathy that could help create brand loyalty and eliminate some of negativity and fear associated with the way humans think and talk about their money, she said.

.. First, we made the deliberate decision to design Eno as gender-neutral. I felt strongly that we should challenge the industry trend of choosing female characters in voice and name for their bots.

Does it matter how humans talk to bots?

I have three young children, and they were the best examples of why we need to be mindful of the conversation we’re establishing with AI. I would watch them in our home interact with [Amazon.com Inc.’s] Alexa. They began by commanding Alexa around and I realized we were establishing these communication patterns that carried through in the way they communicated with me.

.. How does Eno react when a human is disrespectful toward it?

We establish character boundaries for Eno because there’s an insane amount of people who abuse or harass their AI and I didn’t want Eno to just have a generic response. I wanted to establish boundaries and [have Eno] say ‘Hey, that’s inappropriate, let’s stick to focusing on your money.’

We as AI designers have a responsibility to not only the product we put out in the world but in being mindful of those conversation patterns that we establish with our customers. It’s not simple, because there are millions of different ways somebody can say something, and if we misinterpret that and we send the wrong response then we lose that trust.