How to Design User-Focused Chatbot Conversation Flows
Breaking down our approach to building chatbots always comes up. There is a lot that goes on behind the scene of a well-oiled conversational solution.
One of those things is designing conversation flows. Not only that, developing flows that make sense from a chatbot user point of view. When you start to think of the inner workings of a chatbot, you soon realise there is a lot to plan. It's not just 'user says hello and chatbot responds'.
In this article, we'll go through the process we use to design conversation flows. We'll also talk about making sure these are user-focused, i.e. they serve the chatbot's One True Goal.
Conversation flows: the essence of your chatbot
A chatbot is made of many conversation funnels.
So, you had better be sure you're designing them the correct way. Should you fail to do so, your bot could fail entirely.
Important: With most clients, we run through this exercise together as their chatbot project begins. However, as key individuals within your company, you should be able to follow the plan below and fill out the blanks before you even talk to us (or another chatbot building company, boooo).
To design conversation flows, we start from three key points:
- The chatbot's One True Goal
- The chatbot's core users
- The chatbot user stories
Let's break each down and reconcile them into one beautiful, reliable, and replicable approach.
The chatbot's One True Goal
We are massive proponents of the 'One True Goal' approach to chatbots. We use it all the time and wouldn't have it any other way.
The idea is, each chatbot needs to have a One True Goal. This One True Goal is why the chatbot was built. It is the single purpose of the whole thing.
For example, let's take an HR chatbot. Your company needs an HR chatbot. Great. What it is its One True Goal? Is it:
- To reduce the burden on HR staff?
- To make our millennial workforce happier?
- To onboard interns?
There are many ways to build an HR chatbot. Your bot's One True Goal defines why we're doing it, what the end goal is.
When designing conversation flows, knowing the chatbot's One True Goal is crucial. It is the end point of each conversation flow we'll ever design.
Grab a whiteboard and write the OTG (that's right, got bored of writing it) at the bottom in big bold letters.
Everything you'll design will point at the OTG, keep it in mind at all times.
The chatbot's core users
Now we know the end goal of our conversation flows, we need to define who's going to have them. A chatbot is, obviously, a tiny bit useless without users.
Who is going to use your chatbot?
I recommend limiting yourself to four core users. Four users is already a lot to work with, so make sure you are very very sure you need all of them.
Keep your chatbot's OTG in mind. Sure, your CEO could also use the chatbot to check in on the intern's onboarding process, but is she really a user at this stage of your process? Probably not.
List the core users at the very top of your whiteboard.
The chatbot user stories
The final piece of the puzzle is found in a very common exercise: user stories.
A user story exercise is simple. Take each user you've defined above and brainstorm the actions you think they should be able to take through the bot. For instance, taking our HR chatbot for intern onboarding example:
- An intern should be able to get directions to the cafeteria.
- An intern should be able to use the bot without being logged into the company's online portal.
- An intern should be able to book a face-to-face meeting with their manager.
And so on.
I like to use coloured post-it notes, ideally one colour per user.
Important: it's easy to confuse user stories with conversation flows. You'll may fall into the trap of starting mapping out conversations. At this stage, that's not a good idea. The user story exercise helps you figure out all the different actions a specific user should be able to take. Don't plan the conversations around those actions just yet.
Designing the first chatbot conversation flows
Great. We've got all the pieces of the puzzle, now all we have left is to assemble them. Easy peasy!
Designing the actual conversation flows is a time (and brain) consuming process. At this point, you want to:
- Pick a user at the top of the board.
- Align the chatbot's OTG at the bottom of the board.
- Map out the most direct way for that user to go from the first message to reaching the chatbot's OTG.
- Align the user stories you will definitely need alongside that journey to make it successful.
It's easier to make the first pass at it on paper. Once you have, whip out your favourite design-drawing-diagram-making tool and start going full mind mode (we like realtimeboard).
Got your first draft? Start iterating. What have you missed? Where does the conversation fall flat? If a user gets lost during a conversation, how do you get them back on track towards their OTG?
Designing conversation flows can get intense quite quickly. Here's an example of a single conversation flow for a single user on one of our recent HR builds. This was designed during the discovery phase:
Your eyes (and internet loading speed) are fine. I've blurred the image to protect client information. The purpose of this image is to show you the extent of the branching and how in-depth these single conversations can take you.
Done? Get crackin' on your next user!
Where's the smart stuff?
The next step after we've defined all the above is identifying where the smart stuff goes (you know, NLU, NLP and ML wizardry). After all, no one wants a if-this-than-that decision tree chatbot.
At this stage, we bring in the smart people to spot where the smart stuff needs to happen. It's probably my favourite part. Now we start to see the wheels turning and the proper conversational solution being born. It's a beautiful moment.
Defining the users, settling on a unique goal, mapping out the journeys from start to goal; these are all steps you are going to take to properly design your conversation flows. It may all seem very laborious, but trust me, it is a valuable process.
The earlier you go through it, the better your bot will be.