Designing conversations is not something entirely new. All the way back in my student days, as a door-to-door salesman selling everything from gas to milk, we designed conversations. I was 17 and saving up for a car. I worked for a marketing company in Bournemouth, you know, one of those lively, wacky “people” companies that rewarded aggressive sales performance with bell-ringing, chanting and violent hand-shaking, chest-bumps and beer.

Before we’d head out to our patch, we’d be rehearsing the pitch, working on our tone, our banter, our relationship building, overcoming objections, everything we may encounter in a conversation with a potential customer. We’d imagine those conversations, based on previous experience, persona mapping and planning for the unexpected.

We designed those conversations using numerous data points, we just didn’t realise we were doing it, since in sales, this has been done since the dawn of time.

As we now enter an era of Chatbots, we’re designing conversations for more than just sales. As a result, we have people from different parts of the organisation needing to input to the “creative process”, in a way that is very different to what they’ve ever done before. The roles have changed, or at the very least, shifted.

Due to the inherent early nature of Chatbot development, developers are often tasked with “building a chatbot”. They are given the overall responsibility on behalf of the business and the brand to imagine conversations and build them into this new experience. That’s wrong.

There’s an opportunity, absolutely, and a challenge, for sure, for the techies to get much closer to the business, to the customers and this is something to be wholeheartedly embraced.

We need designers involved. But in a Chatbot, there’s so little graphical UI that the typical graphic-oriented designers mindset is somewhat misplaced. This requires the designer to get even closer to the business, the customers, the UX and this is something to be wholeheartedly embraced.

The Bot Playbook

We need customer support to be involved. But in a chatbot, whilst there’s some room for emotion-based, gut instinct reactions to typical questions, like “what are your most commonly asked questions?”, “how do you normally respond to X, Y, Z?”, it can’t be everything. We need a balance between hard data, derived from call logs and ticket systems along with instincts, tone and emotion of the delivery of world-class customer support. It requires a team to get that out of the system and into a creative process.

We need sales to be involved, we need production to be involved, we need marketing to be involved and anyone else who either has a place in the normal everyday (non-bot) version of the conversation.

The Future of AI Is in the Hands of Storytellers

So, how on earth do we manage this in a creative and production process, designed to build and distribute bots?

I Use a Mural

Mural.co is one of the key tools in my workflow. Basically, it is a map of some messy parts of my brain and my business, in those all important stages of starting something new — an idea, a new task, a problem to solve.

In this context however, it’s a great and super easy way to collaboratively design conversations with a team, in pursuit of the goal of building awesome bot experiences.

All of this can be done in the real world, with post-its. But I’m a distributed teams kind of guy, so I prefer online collaboration software.

The Legend

These are the basics of our design process, and key terms within a conversation design process aimed at producing bots.

User Says… the yellow post-it is the user input, simply.

Bot Says… the green post-it represents the Bot’s input. Easy huh??

Intent… the orange post-it represents a prediction of what the User is likely to be intending to say or do, based on the input and taking into account the current Context of the conversation.

Context… in the purple post-it, the context is a kind of state, which enables you to manage and allows the Bot to more accurately estimate what the User means when they say something.

Param Group… this is information that we need within the flow of a conversation, to complete some task. Think of it like a form, which gathers information, but in the format of a conversation using questions and answers. The blue post-it represents the parameter we need, and if it’s not offered by the user voluntarily in the natural course of a conversation, then we can request it by way of a prompt, the green post-it.

Action… the pink post-it represents some action that needs to be taken by something. Imagine you’re talking to someone, instructing them how to make a cup of tea, then an action could be a trigger that tells that person when to boil the kettle, or take out the tea bag. It could also be a trigger to go online and order some more tea bags when you’ve run out.

Understanding how all of these elements work together is the key to designing conversations. Enabling all the people in the team to contribute to the process in the right way, is the secret sauce.

Designing Conversational UI with Information Architecture, Part 1

Mural makes it very easy to sketch out the conversation, say, in terms of the various User and Bot inputs. This is the heart of the conversation, and typically something that anyone can have a stab at. But only when those who represent the usual voice / tone of the business get involved, do these notes reflect a useful brand communication proposition. Otherwise, at this point, it’s just a toy.

From here, the designers and developers (and in many cases, our next-generation super-developer / super-designer (another post!) who has an equal grasp of both domains) can start to attribute the intents, contexts and actions.

We can prototype the naming conventions for these, which will undoubtedly grow beyond the ability to manage them without careful consideration.

The conversation can grow organically, as would a normal conversation. New topics, new branches and threads can sprout up. New contributors can join in and create new conversation flows. Each aspect can be commented and discussed, to optimise, optimise, optimise.

I find it makes for a very healthy, simple and cost effective creative workflow that requires people to contribute effectively as we all continue to figure out how these things need to work.

I encourage you to give it a try.