The bots are coming! It seems every other day sees a new article discussing the possibilities of conversational user interfaces. Brands such as Tech Insider and CNN have created their own versions to put content in the hands of their customers while KLM, the Dutch airline, is using one to enhance their customer experience. Still in their infancy, these conversational services are creating big waves in our industry.
We’re incredibly interested because of the endless possibilities. Human language is nuanced, diverse and amazingly intricate having been developed over thousands of years — meanwhile, gestures, touches and other screen-based interactions can still feel unnatural and limiting within the possible experiences that can be created. Will Chatbots transform the way we deliver products and services in the coming months and years?
We find the possibilities fascinating, and it let us to think how we might create a compelling customer experience for our clients using a chatbot. Here is what happened when we paused the research and instead started experimenting…
What we did
Using a platform called ‘Chatfuel’, we built a bot that lived on Facebook messenger. Our goal was to help users learn about one of our client’s products, allowing them to begin defining their needs and compare similar products. Our process was to run a sprint where we built, tested and iterated our bot alongside users and designers.
Understanding the audience
One of the very first things we found was that our chatbot fostered an experience that the customer could dip in and out of as frequently as they wished, changing the typical interactions they would have with a brand. With this in mind we had to understand their needs, intent and every possible twist and turn they would take. We had to know how this service influenced both digital and physical discovery of our product, purchase and later steps of the journey.
We have an ethnographic approach that we use in order to gain a comprehensive understanding of user needs, we conduct interviews with key stakeholders, observe the journey in-store and online, take part in experience walk-throughs, and use research kits that are given to users to complete within context. All of these qualitative findings are synthesised until we have a clear picture of the needs and expectations of the brand throughout the user’s journey. Without this approach we would find it extremely difficult to know how to tell a story to customers or inspire them to action.
What are the new interactions?
Considering how early in their adoption chatbots are, both customers and designers are yet to become familiar with the potential interactions and experience. As designers, we can’t draw directly on much of our previous experience with interfaces, the patterns we are familiar with don’t work — it’s more difficult to prompt a user’s experience and draw them towards a chosen path. For example, Amazon Echo has no screen-based interface, this puts the emphasis on the user themselves to drive the interaction. It means they must know how to open the discussion and how to respond. This potentially risks embarrassing users, and might lead to them ending their experience. To reassure users, Echo responds with a visual light to show that it is listening, meanwhile Apple’s Siri replies with a ‘let me think…’.
How we came to afford interactions to users became extremely important, we had to show them that we were listening and working on an answer. Interestingly, we found the speed in which the bot replied to users could reduce the perceived value of the result. We didn’t want users to feel as if there was a person on the other end of the conversation. However, we found that by delaying the results just a second or two led to a perception that their query had been considered, and they felt as if the bot had found them a worthy answer or the best deal possible. Ultimately, experimenting with delays made the whole experience more believable to users.
How did we tell our story?
How we phrased something has become more important than ever before. We found that the positioning of the bot’s capabilities had to be super clear, and (if it must) gradually become more complex. When we didn’t do this, we risked overwhelming the user to the point where they wouldn’t know what to ask the bot to do next.
We realised we had to quickly become great storytellers. Once we understood the user’s end to end experience we then mapped out each step. Through it we matched the emotion we wanted to evoke in the user and experimented with what we wanted to tell them at each of these points. The rigor and detail in this process allowed us to create something that was compelling and fostered moments of delight throughout the journey.
As part of the storytelling method, we had to define a tone of voice for our bot. Our experiment was for one of Wunderman’s clients, and whilst they have a clear tone of voice in their current communications, we had to define what their conversational tone would become. Like many existing bots we experimented with levels of personification, at times it became a little creepy, weird or even fun. We felt the best way to discover what worked was to iterate fast, trying different things and seeing what did and did not resonate with users.
What did we learn?
Now, looking back on our journey of creating our first chatbot we’ve had a think about some of the decisions we made and what we learnt along the way. Hopefully these will help us, and you, discuss future experiments with bots.
1. Know your user’s existing journey and needs
While fundamentally changing this journey, you need to know what you can influence and in what moments your chatbot will exist.
2. Know what emotions you’re trying to invoke
By understanding these needs, you can apply new emotions and ensure your bot provokes them.
3. Making a user wait can actually improve their experience
Our experiences with these products are new and users need reassurance during their interactions, build appropriate affordances.
4. Focus on gradual discovery over complexity
We’re yet to discover everything we can do with bots and so are users. By gradually surfacing new features we are less likely to overwhelm our users.
5. Favour tone of voice over personification
We found that personification didn’t resonate with users during our experiment. Instead, having a distinct tone of voice was the brand distinction. We love ‘clippy’, but we don’t want to be him.
6. Allow users to lead the conversation
Chatbots fragment the typical user journey, to ensure we answer customer needs we need to allow them to lead the conversation. We should keep up and be ready to change topic.
7. Focus on the ‘little victories’, the tiny moments your bot can influence.
Don’t try to hold a complete customer journey in a chatbot, instead supplement the existing journey by being there when you are needed.
We learnt a lot from the process and the bot we created, however we are far from finished. Conversational interfaces are still in their infancy. Moving forwards we’ll need to keep experimenting with what works and what doesn’t for users, but this process will allow us to create more exciting opportunities than ever before. If you take one thing from this article, please, go try and build your own. If you take on the challenge, we would recommend Chatfuel for making it super easy to iterate and test what you build.
Credit to Alex Zeeman for the header.