How character and design blend together to build the perfect Chatbot

Read full story
Growing up in 70's Kolkata, I remember the frequent blackouts in the evenings. Not with disgust. But with fond memories. It was a signal for us to break into stories and songs. And we almost reluctantly blinked at the light when it returned. "You're never going to kill storytelling, because it's built into the human plan. We come with it," said Margaret Atwood. Is that as true today as it was years ago when we had time for conversations and stories?

And does it hold true in business? Customer Experience is the big battleground today for businesses looking to enhance customer retention and engagement. But whether we are interacting with a customer service representative, filling up an online form, or communicating with a chatbot, can technology erase our human love for storytelling and conversations?

Bringing back the conversations - a chat with a chatbot:
No buttons, no menus. Just a simple conversation with a customer. That's a promise of a chatbot. Chatbots are not just growing at a CAGR of over 30%, they are also aiming to bring back the art of conversation to technology driven customer experience solutions. Is it that simple?

A chatbot is supposed to represent a conversational interface with text, voice, images that can sound just as a real conversation would. So, what goes behind creating these conversational interfaces? And what keeps the conversation real? The obvious answer is AI and machine learning. Like every new product, chatbots go through their own design, development and deployment cycle. They are trained with data to learn from human language and develop a vocabulary of their own. But is that sufficient?

Look at this chatbot for example. I decided to test out its promise helping discover every aspect of life through a poetic touch, a tough ask for even AI and machine learning. I quickly realized that the chat's definition of being poetic was just sharing quotes. And not even original ones, at that.

What was really jarring for me, was the lack of character and design in the chatbot. It was clear that I was really not conversing, just getting quotes that I could have downloaded from Goodreads.

Building a good character - the chatbot way:

A successful chatbot is just that. A good conversation. Just as in a fiction story, what makes it work are characters that are memorable and consistent in their quirkiness and not just cardboard cutouts. Imagine Huckleberry Finn suddenly turning polite. Or Scarlett O'Hara from Gone with the Wind becoming unselfish. That wouldn't be "in character."

Are characters for chatbots built the same way?

Jagrat, UX designer at Haptik, one of India's highest rated chatbots says, "creating a chatbot is much like creating a script for a conversation between two actors. One actor you know because you created her -the chatbot- and the other actor is the unknown consumer. And just like a writer, we have to create an imaginary world where these two characters can have a real conversation. A chatbot has to develop a distinct persona which sets the tone of the conversation."

Turning up the Tone volume:

Let's take two very different examples to understand this. Poncho and Haptik. Poncho shares weather and traffic updates, Haptik solves specific customer queries on setting reminders, travel requests, bookings etc.


Their characters? Poncho is quirky, humorous, opinionated. Haptik is dependable, helpful and fast.

Here's how the tone of the two chatbots differed when I started conversing with them:

Sipping my cup of hot tea on a rainy Bangalore evening, I asked Poncho for a quick weather update. Poncho gave me the information I needed and added some irreverent opinion of its own. The tone was a mix of informality and humor.

I asked Haptik to set up a simple exercise reminder for me. Haptik first asked me what I needed - the date, the type of exercise, the time and then confirmed it. Polite, dependable. Fast.

Both the chatbots were in character. And they have to stay in character right from the welcome message to the error message. Enter, Design.

A designer smiley


A single smiley can make or break a conversation with a chatbot. Bad design can kill it completely. When I move from simple reminders to more complex tasks, it is easy for a chatbot to lose context. Imagine asking a chatbot questions like "Can you web-check me in for the flight you booked for me two weeks back?" Machine learning may not have the answers. And content alone may not be enough.

Haptik segregates "flights" from "all travel" to reduce chances of errors and uses forms or lists to encourage more tapping. These make the conversation crisper and tighter. I tried typing a flight number without telling Haptik what I needed it to do. Rather than giving up, Haptik gave me a list to tap on. Less typing for me to do. And a clearer query for Haptik to solve.
The designer on the other hand has a lot more work to plan this in advance. Use of carousels, sliding photos, typing indicators. Even down to the length of time between the messages and the colour on the screen.

Contrast this with what I found in Poncho. As I moved from weather to traffic updates, I met an unexpected bottleneck. Traffic updates were not covered for my location, Bangalore. The colors and the font sizes were bolder in Poncho. But the message was still clear. Plain, simple. Poncho could not solve this query. No room for confusion.

Beyond the chat-ter:
There is endless chatter around the future of chatbots. Will generic platform chatbots seize the day? Or will branded chatbots become more pervasive? Will a hybrid approach of chatbots along with human interface win the art of the conversation? Over the next few weeks, we will explore each such aspect of the chatbot's journey in more detail in a series of articles that go beyond the chatbot chatter.

(This article is part of a 3-part series on chatbots brought to you by Haptik and Business Insider. It looks at why chatbots are growing the way they are, who chatbots can help and what they really are capable of doing alongside highlighting the key opportunities and challenges. At the end of the series, we will also provide a consolidated and detailed report that you can download.)
{{}}
Add Comment()

Comments ()

X
Sort By:
Be the first one to comment.
We have sent you a verification email. This comment will be published once verification is done.