On Bots, AI and Content Strategy

Cruce Saunders This is "Towards a Smarter World," and I'm your host Cruce Saunders. Very pleased to be joined today by Elizabeth McGuane, who is the content strategy lead at Intercom where she is part of the product design team, and owns the language of the core product including its messenger app. Elizabeth's been working in UX for 10 years and before that was a journalist. I'm really glad she could be with us today. She recently wrote an amazing article to check out on TechCrunch, called "On Bots, Language, and Making Technology Disappear." Elizabeth, with that article, would you summarize some of your thinking behind how you ended up arriving the conclusion that actually naming a bot is not necessarily the best strategy? We did it through research, but I think where we started was through a really careful and considered approach to testing the language. When I started, this was one of the first projects I worked on at Intercom when I joined just over a year ago, and we were looking at introducing bot-like, very simple bot, into our messenger. We make a B2B messenger, so not to get to complicated in terms of the UX of our product, but we always have to think of our users in terms of two layers: we have our customers and then our customers' customers, and we were really creating a bot that businesses would use to communicate concepts or to get data from their customers. I knew that we need to be really careful about how we express things so that we would marry with the business' tone of voice so that we wouldn't be overstepping the bounds of what we could say on their behalf. I had a feeling, and this was really just my gut instinct, that having a very chatty personality would not necessarily marry with the tone of voice of every single business that wanted to use our messenger. It was a very practical consideration on that front. When we went into testing, we tested with a name and without out a name. We also did testing with different tones of voice because going into this I think the design leads were interested to see whether a more friendly tone of voice or a more functional tone of voice would work. That was the initial consideration of "let's just try different kinds of copy, and see what works." I felt that I wanted to take a more structured approach and try names, no names, functional, more friendly, then we also tried with a pronoun, without a pronoun. Once we realized that names didn't work we also tried removing the first person "I", and removing an introduction so that the bot didn't say, "Hi, I'm so-and-so's digital assistant," or what have you to see what impact that had. That's really where it started was with an actual structured approach to research.