Latest news

Designing a Conversational User Flow for Useful Bots

No comments

3 Questions. 1 Flowchart. Infinite Possibilities.

This animation was created by Michal Ptaszynski via Dribble

Conversations Take Unexpected Turns

You might have an overall topic you’re chatting about, but being prepared for anything is something inherent in human conversation and something radical when you’re designing artificially intelligent (AI) experiences.

Unlike more conventional interfaces, the conversational UIs that lay on top of artificial intelligence in bots need to adapt to anything said in conversation. So there’s this delicate balance of mapping out a path and also inviting unexpected input. In this piece, I’ll walk you through some of the steps I take in designing a conversation for The ShaBot*.

Every Conversation Designer Should Ask Themselves These 3 Questions

To tackle the big picture goal of designing a personalized and dynamic chat experience — here are 3 questions to keep in mind.

  1. What’s the user’s goal?
  2. What’s the bot’s goal?
  3. What can you do about unexpected input?

What’s The User’s Goal?

To nail down the user’s goal it’s valuable to talk to real people and ask what they think. I started by chatting about the idea with friends and then dropping a survey to anyone who liked the page.

the survey I posted in March

The data I gathered led me to radically change direction. Instead of trying to fill the bot with cool features, I narrowed focus and chose one problem for the ShaBot to solve at a time and that was based on a goal that a number of people asked for — how to learn when the sun sets in the simplest way possible.

In sum, focusing on one goal and validating that that matches the user’s expectations is an essential first step to making your brand accessible (bots are fairly new and education is a key first step).

Once I knew the user’s goal, it was time to design the bot’s goal.

What’s The Bot’s Goal?

The bot’s goal at each stage is to help the user achieve their goal. The ShaBot uses Facebook Messenger, so those constraints affect the user’s experience and afford the usage of natural language processing (NLP) to process user input and deliver the intended text output or buttons.

an example of Messenger’s native buttons used to respond to a user’s question

Taking a step out of the context to map out the user’s journey is helpful in scoping out the entirety of the bot’s functionality. I’d suggest starting at the highest level with a few branches that display the simplest path with space for any input (without prescribing a response for any input).

A User Flow for The ShaBot with space for any input here scoped out in yellow dotted lines

What Can you Do About Unexpected Input?

Once you’ve designed a structure that would serve your user’s goal at a high level, it takes reading into how each person uses the bot to develop granular responses to direct users towards their goal in a conversational way. In other words:

“Your words are raw data that teaches us what you want from us”

Steph Hay, Content Design at Capital One

Simply put, observe your users closely. Then, design AI driven responses that acknowledge user input and reiterate the bot’s goal. That takes time, and could mean many many tree branches to capture everything. So, for now, I like to focus on the big picture goal and then drive personalized responses that push that goal back into view. Designing a bot that listens starts with actually listening!

Here’s a fun example I discovered in the ShaBot’s transcripts. Someone asked for something and then thanked the bot!

my bot didn’t know how to handle a thank you! (via a meme of Nick Kroll)

If your bot is awesome, people might respond with delight! Input could come in any form, but in this case it was positive. So I built a bank of responses to any appreciative input moving forward. #failfast

improvement to the content based on reviewing input

It may seem obvious, but to improve on previous conversational experiences it’s important to review each one— especially the painful ones — and take action to support your user in achieving their goal while actually listening and responding to what they say.

Summing it up

Designing a great conversational experience starts with thinking less about what you want to say and more about what the user wants to do.

How do you structure conversational flow?


*The ShaBot is one of my side projects, and it’s like my sandbox for trying out new ways to use the conversational medium. It’s a simple chatbot that lets you check when the sun’s setting near you on Friday, which is the signal that Shabbat — here referring to the Jewish Sabbath — has started. I like to think of it as a lighthearted blend of ancient tradition and brand new tech. Got questions? Ask The ShaBot!

Designing a Conversational User Flow for Useful Bots was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Source: Chatbots Magazine

magnoliaDesigning a Conversational User Flow for Useful Bots

Related Posts

Leave a Reply