Since we first started building computers, humans have gone to great lengths to interact with our machines. We’ve graduated from punch cards to keyboards to touchscreens. We’ve written code into command lines and learned a thousand icons. Useful as they are, digital systems haven’t been particularly easy to engage with.
Conversational user interfaces change all that. They place the burden of understanding on machines, not on people, by allowing us to enter commands as naturally as we greet our families in the morning. As a result, they create more engaging experiences for users.
Today, companies are designing voice user interfaces for a growing list of applications: chatbot-guides on websites; interactive voice response systems for AI-driven call centers; voice apps that bring brands into the audience’s living room; even smart devices for the burgeoning Internet of Things. But not all of these conversational UIs provide a great user experience, and a frustrating tool can be a disaster for any brand. If you ever find yourself contributing to a conversational UI design, here are a few things to keep in mind.
How Conversational UI Drives Value
Before we get into tips about how to build a great conversational UI framework, it might help to understand why conversational UI is so attractive to companies. The specifics depend on company goals, of course, but investments in conversational AI can open up new shopping channels, improve existing digital sales, vastly reduce customer service expenses, and provide more convenient customer experiences than ever—without hiring whole new teams.
Voice user interfaces are also powerful branding vectors, provided companies invest in custom branded voices that establish them as familiar and authoritative presences in their audience’s lives.
Defining Conversational User Interface
A conversational user interface, or conversational UI, is a system that allows people and computers to interact with everyday language. That language may be written, as in a chatbot, or it may be spoken, as in a voice user interface (VUI).
Conversational user interface examples include both VUI and chatbots. Voice assistants like Amazon’s Alexa, Google Assistant, and Samsung’s Bixby are familiar examples in the VUI category. Chatbots are popping up everywhere, often with great success. Activision’s Call of Duty chatbot racked up 6 million engagements on Facebook Messenger in its first week alone. Another chatbot that made headlines belongs to floral disrupter 1-800-Flowers. The company’s bot served as a sort of proof-of-concept on the Facebook platform, prompting Mark Zuckerberg to quip, “To order from 1-800-Flowers, you never have to call 1-800-Flowers again.”
As these conversational UI examples illustrate, the category is broad. But all conversational user interfaces have one thing in common: artificial intelligence. The technology behind conversational UI is a subset of AI called natural language understanding, or NLU. With NLU, machines extract user intent from a wide range of utterances, regardless of accent, phrasing, or background noise.
6 Best Practices for Conversational UI Design
Building a great conversational UI experience requires two things: Designing the right thing, and designing the thing right. In other words, make sure your bot meets a real customer need. Then build a system that works reliably. These tips will help you do both.
1. Start by identifying specific use cases.
To serve customers, you have to identify their pain points. Only then can you build a solution that people actually want or need to engage with. So the first step in conversational UI design is to identify use cases, i.e., the thing your UI will enable your user to accomplish or the need it will meet. When you choose your use cases, you’re essentially deciding what your conversational UI product will do. And they tend to stem from your company’s aims and goals.
Imagine an insurance company that needs to reduce the number of calls coming in through a contact center. They may identify use cases by separating calls into categories, then identifying the category with the highest volume and the lowest complexity. Say they get a lot of calls simply asking about policy charges; this would be an ideal problem for a voice bot to solve.
That’s finding use cases starting with value to the business. Another approach is to start with the user. Construct a user persona based on industry data and company research. With that persona in hand, walk through a day in their life, from opening their eyes in the morning to going to bed at night. Pick out the pain points where your bot could help. Those are clues as to what you should be building with your conversational UI.
2. Validate your choices before building anything.
Once you’ve narrowed your list of use cases to the few that could provide the most value, you’ll need to test them. One way to test potential use cases is to role-play a conversation, with one participant playing the user and another playing the bot. Another option is to use actual transcribed conversations, if you have them, as many call centers do.
Walking through conversations will help determine whether use cases are realistic. They’ll also provide more details about potential conversational digressions you’ll need to plan for or technical requirements to consider. You’re looking for what conversational UI designers call the Happy Path—an ideal conversation between the user and the bot, one that lets the bot accomplish its goal without distractions or tangents. Validation exercises help identify both the Happy Path and the potential ways speakers may veer off of it. Both are necessary data for building your bot.
3. Learn more from a prototype.
With validated use cases, a Happy Path conversational model, and a few common conversational digressions identified, you’re ready to take the bot off the page. A prototype lets you interact with your system, trying different conversational approaches and zeroing in on potential failures. Early work with a prototype will lead to what developers call an interaction model.
The interaction model is the guts of a chatbot. It includes all the things people may say, the intent behind those words, the system’s responses in each case, and the specific data the system needs from the speaker to fulfill a request (those data points are called “slots” or “entities”).
4. Get end users involved in the early testing phase.
Invite end users to try out the prototype whenever possible. There’s no better way to reveal the surprises that may unfold when people interact with the finished product. The more users you can involve in testing your system, the better.
The goal of all this prototype testing is to create an interaction model that’s as robust as possible. Along with technical requirements, this is what you need to start the true software development process.
5. Prioritize functional use over speed to market.
There’s nothing wrong with a minimum viable product (MVP), the quickest, cheapest version of a tool that’ll meet your core user needs quickly. Sometimes MVPs provide value. That’s the idea, but they’ll only provide value if you get your use case right. So we’re not saying never hurry, but we are saying be careful. If you have to choose between getting your use cases right and quick deployment, finding the right use cases should win every time.
If you rush, you may end up with a bot or a voice assistant that does not provide any value to the user. That leads to disappointment and can even damage your brand. More often than not, the carefully considered bot that takes longer to produce and design properly will provide more value over time than the rush-job that’s first to market.
6. Make sure your bots reflect your brand.
Chatbots and conversational interfaces aren’t just tools; they’re also important branding opportunities. After all, they interact directly with customers. Structure your chatbot’s responses to express a memorable personality and give them an identifiable writing style where appropriate. Voice user interfaces should use a unique text-to-speech voice, and that voice should closely match specific brand traits.
One thing’s for sure: Chatbots, voice assistants, and other conversational UI can be powerful touchpoints between brands and consumers—but only if you build the right thing and you build it in the right way. Follow these tips to make sure yours does.
Kane Simms is CEO of VUX World, a consultancy and design firm that uses conversational AI to help companies boost revenue, cut costs, and improve customer experience. Take the VUX World free voice strategy assessment to get your conversational AI strategy started In the right way.