Chuck Gose: You’re all over the place lately. I thought you were being affected by the hurricanes, and now I learned you’re back up in America’s hat. . . Canada.
Chris McGrath: LOL. Yes. We’d been in the Caribbean for so long that my kids had forgotten all about the country on their passport. So we’re doing a ski season and school year in Canada, and then after that, we’ll see...
CG: Caribbean. . . skiing. . . rough life for Chris.
CM: Ha. Well, early on I started taking advantage of this remote working thing, and it’s enabled us to do some interesting things.
CG: How long were you living in the Caribbean?
CM: 6 years in St. Lucia. The little guy can’t remember snow; he was 3 when we left.
CG: I wish I didn’t remember snow.
You know, using LOL all the time is something I picked up in the Caribbean. It’s all abbreviations and emojis down there...
CG: That’s super interesting. Why do you think that is?
CM: Regular phone calls are too expensive for the average person, so they adopted texting very heavily. But that was too expensive for heavy texters, so when mobile internet became available it was adopted swiftly, mostly so people could use Whatsapp. That’s now the de facto way to communicate with everyone on the island: Whatsapp.
"Chatbots require a lot of writing and content work. Internal communicators aren’t getting off the hook here!"
CG: That’s a nice segue to the work you’re doing with Tangowork. For those who aren’t familiar, explain your creation.
CM: It’s a software service that powers chatbots for internal communications. It runs inside your regular messaging app, like Skype or Slack. Ask it questions about company stuff using regular language, and it answers back.
CG: Do you have any information on the growth of chatbots for all uses, not just internal comms?
CM: Facebook releases numbers for Messenger chatbots on a semi-regular basis, so that’s about the best information I have. Earlier in 2017, there were about 100,000, growing at roughly a 300% annual rate.
CG: Recently you were in London presenting at the Intranet NOW conference. What were some of the questions or feedback you received at the event?
CM: There was a ton of interest in my presentation, and then I was swarmed during the roundtables with questions about chatbots. I met some people who are actively building chatbots for internal comms, and others who had heard of chatbots but had never thought about their application internally.
"That’s a key point of differentiation with intranets, the other common internal comms vehicle. Intranets are pull-only, whereas chatbots are both push and pull."
CG: What are some of common misconceptions communicators have?
CM: I’ve been asked if the chatbot can just be installed and then automatically figure out what to say. That would be wonderful! But that’s not really possible. Chatbots require a lot of writing and content work. Internal communicators aren’t getting off the hook here!
CG: Do you think chatbots should be reactive, proactive or both?
CM: Both. A chatbot responds reactively to user input, of course. But there’s also an opportunity to send proactive notifications via chatbot.
For example, if a weather event is about to affect your business, sending a proactive chatbot message, received on your employee’s phone, is likely a faster way to get in contact than email.
CG: That’s a misconception, or perhaps misunderstanding is a better word, that I think internal communicators have about chatbots. It doesn’t just have to be sitting there, waiting to receive a message. It can initiate communication as well.
CM: Yes. That’s a key point of differentiation with intranets, the other common internal comms vehicle. Intranets are pull-only, whereas chatbots are both push and pull.
CG: I read a really interesting article on LinkedIn about Capital One’s chatbot. It has nothing to do with internal comms, but I thought it had a lot of great lessons learned. I thought one of the cooler things was that Capital One learned how its customers were texting and what words they use.
CM: I’ve done a lot of user testing over the years, and it’s always incredibly insightful.
But most of us don’t do nearly as much testing as we should. And what I’ve discovered about chatbots is you’re doing constant user testing. Every conversation transcript is like a user test. When you read through the transcript, you see what the user is thinking and trying to accomplish, and then you see the chatbot either succeed or fail.
For the last two days, I’ve been studying transcripts from a new client launch, and it’s just amazing what you learn. People say things you never could’ve imagined, and then you adjust the chatbot to handle it. And when the chatbot absolutely nails a conversation—when it’s a perfect transcript — you do a fist pump and pat yourself on the back. It’s really a lot of fun.
CG: It’s probably a bit like eavesdropping on a bunch of conversations.
CM: Yes. It can be a bit embarrassing, especially when the user is hitting on the chatbot.
CG: Whoa, that escalated quickly. Have you seen where employees think they’re chatting with a real person?
CM: No, so far everyone’s clear that they’re talking to a computer. But I think that’s why employees may feel emboldened to say things that they’d never say to a human.
It’s also a side effect of giving your chatbot some personality. Because chatbots are dealing mostly with text, there’s very little design or branding in the traditional sense. So the personality becomes the brand, and if it’s well done, people connect with it on some level.
CG: People like to name channels - their intranets, especially. Should chatbots have a name and what are some clever ones you’ve seen?
CM: We have a personality workshop that we use to help the client team make decisions about personality. After looking at some examples, we decide: first person or third person? Name or nameless? Gendered or genderless? Human or computer? Understated or enthusiastic? And so on.
I like chatbots to have a name, even though I know they’re a computer. It just seems so much friendlier.
"It’s easy to get distracted with the [chatbot] personality when that’s really the final consideration."
CG: The Cosmopolitan of Las Vegas has a chatbot for guests. Her name is Rose. And she’s just as saucy as their brand is. Are chatbots a way for an internal brand to expand a bit or do most like to stay in their safe space?
CM: There’s a hierarchy of needs you need to consider with chatbots, as with other applications.
First, it must be functional. The thing has to work.
Second, it must be useful. It must help me with my day-to-day job.
And finally, it should be delightful. And that’s accomplished with the personality.
It’s easy to get distracted with the personality when that’s really the final consideration. It’s a ton of work to get the chatbot to “functional” and “useful.”
CG: I like to get distracted.
CM: I think all of us do. I love working on the chatbot personality. But a chatbot can be very successful with zero personality, so long as it’s useful. And an exuberant chatbot that can’t answer questions properly is off-putting.
CG: Do you find particular types of organizations are attracted to chatbots and their possibilities?
CM: Initially, I expected that companies with a non-desk workforce would be attracted to chatbots. Retail, logistics, manufacturing, etc.—they have all these employees walking around, messaging apps in their pockets, and it’s a perfect way to reach them. But in reality, it’s been companies with a high percentage of knowledge workers that have shown the most interest.
CG: Measurement is something that internal communicators should all be doing, but it has also been a struggle. What kind of measurement exists for chatbots?
CM: Several third-party analytics packages exist for chatbots. However, a lot of the data that is tracked is more relevant for external chatbots than for internal ones.
We’ve been working hard at expanding our analytics capabilities at Tangowork, and our biggest area of focus right now is on measuring the chatbot’s success rate. On the initial pilot, the conversation success rate is typically quite low. As the pilot continues, the success rate steadily climbs until it’s high enough for wide launch.
That’s a key metric for chatbots: what is the conversation success rate?
CG: Earlier you were talking about the content work that goes into building a chatbot. What role does or will AI play in the success rate?
CM: AI means different things to different people. The part of AI that is instrumental to chatbots is natural language processing. During pilot and afterward, we do supervised machine learning to train the chatbot on how to respond to things it hasn’t heard before. It’s instrumental to the success rate.
As time goes on, we’ll have larger datasets to pull on, and very little training will be required for common queries. But for queries that are specific to the company, there will be quite a bit of work involved, for the foreseeable future.
CG: From what I’ve seen on the IC side, chatbots have been focused on text. Do you think voice will play a role down the line?
CM: In certain industries, yes. For example, if you work in a lab, it could be really useful to just call out, “Hey chatbot, what’s the melt temperature for nickel?” and the chatbot shouts it out to you. And you don’t have to take off your gloves or anything.
CG: Kind of like how my kids have figured out that “Alexa” can help them with their math homework.
CM: Exactly. But in other environments, it might not work. It could be weird to talk to your computer in a sea of cubicles. But maybe cultural norms will change. A decade ago I thought it was weird when people were walking down the street looking down at their phones. Now my neck hurts too.
"Within 4 years 50% of enterprises will spend more on chatbots than on mobile app development."
CG: Tangowork was the first company I saw that was focused on chatbots for internal communication. But I’m sure you have other competition and that will only increase. How do you differentiate?
CM: At this point, the focus on internal communications is our differentiation. It’s a massive addressable market out there — Gartner is predicting that within 4 years 50% of enterprises will spend more on chatbots than on mobile app development. So there will be lots of work for everyone in this space.
CG: I already know you’re a fan of emojis so that last question will be an easy one. And I have a guess what your answer will be. Describe your thoughts on IC via emojis.
CG: I would have guessed 🤖 .