Voice UI Prototyping: A 6-Step Process

Technology moves fast—so fast that many users get left behind. Each new generation of computing requires us to develop new skills, and yesterday’s power-user can easily become today’s Luddite.

But we don’t have to ask humans to constantly update their tech skills. Instead, we can make tech more human. Voice computing is a powerful way to accomplish this goal.

“Voice is the next step in bringing technology to the human level, making sure it’s really there for everyone. It opens up so many possibilities for so many more people.” — Tess Tettelin, conversation design lead at Chatlayer

Voice computing doesn’t require manual dexterity or advanced computer skills, so it reaches a broader audience than keyboard- or touchscreen-based systems. And voice user interfaces (VUIs)—which mediate human interactions with computer systems through voice or speech commands—are available to organizations of all sizes today.

Most consumers are familiar with VUIs, too. They use them to interact with voice assistants, like Siri or Alexa, or to operate smart speakers. Voice controls are showing up in ambient computing devices from smart home systems to wearables. In fact, about 50% of consumers already use voicebots and voice assistants in their day-to-day lives.

Thanks to today’s conversational artificial intelligence (AI) tools, anyone can build a voicebot, smart assistant, or similar VUI application. That bot-building process begins with prototyping.

So how do you build a VUI prototype? We asked Tess Tettelin, conversation design lead at Chatlayer, how voice UI prototyping works in practice. Here’s what we learned.

Voice UI Prototyping: 6 Steps and Tips for Success

Follow these six steps to build a voicebot prototype that will provide terrific results in the field.

1. Define the problem you want to solve.

Voicebots are a means to an end, not the end themselves. Conversational AI systems accomplish specific goals. So the first step in voice UI prototyping is to define your use cases.

Tips for Choosing Voice UI Use Cases

“When it comes to use cases for a voice UI, it’s best to have short or repetitive tasks,” Tettelin says. Voicebots excel at step-by-step information, answering common questions, and—for device controls—performing a clearly defined act.

Also look for “use cases where the bot has to be engaging, and the focus is on creating a relationship or a bond,” Tettelin says. “For example, online therapy, bots that read stories to kids, or HR performance reviews. Our brains are wired for talking. It takes less cognitive load to talk than to type, so people will give more context in speech. So if your use case requires a lot of context, it’s nice to do it over voice.”

2. Know your audience.

Identifying your use cases is a good start—but to design a great prototype, you also need to know who will be using this voice UI. Create audience personas to narrow your audience down to specific users. That will help you design systems that address personalized goals.

Tips for Identifying Users

“Any business will have the obvious users, the people you think about straightaway,” Tettelin says. “But don’t forget about the users who are not so obvious.”

Think outside your typical audience age group, Tettelin recommends, and consider all your users’ accessibility needs.

“You might serve people with physical or vision impairments,” Tettelin says. “When building a voicebot, do not forget about these people. Remember: Not everyone looks like the people on this bot-building team.”

3. Create a voicebot persona.

Conversation is personal; it is very human. And people respond better to conversational AI systems that have a consistent identity. Create a persona that matches—and enhances—your brand identity. Then choose a synthetic voice that expresses that identity clearly.

Tips for Establishing VUI Identity

Brand voice is a question of both language choice, voice quality, and expressive speaking style. “If you’re building a bot for kids, it shouldn’t be too strict or formal,” Tettelin says. “It should be loose and use certain words that kids use.” That’s an example of language choice.

Voice quality and speaking style are also crucial to establishing your bot’s identity.

“A bot with a soft voice, that speaks a bit slower, immediately sets a certain mood,” Tettelins says. “A voice that’s really enthusiastic is different, but these differences define your persona. It’s important to be purposeful about the bot’s identity, because when people hear a voice, within 10 seconds they’ll make up their minds if they like it or not.”

4. Script conversations for each of your VUI’s actions.

Once you know your use cases, your audience, and the bot’s persona, it’s time to start writing. Imagine the questions or commands users will bring to the system. Write helpful replies. Then imagine how the user might reply, and map out these various conversation flows on the page.

Tips for Voicebot Conversation Scripting

Conversation scripting starts with the ideal interaction, then branches out from there.

“We have this thing called the Happy Path,” Tettelin says. “That describes an interaction where everything goes according to plan; the user is following the script and everything’s going well. But then you have Edge Cases, where a user might ask a very specific question, or they have a very specific need, that veers off the Happy Path.”

For instance, if you build a customer service voicebot for a restaurant, a standard delivery order might be your Happy Path. But what happens when a caller asks if a dish is gluten-free?

“Most users won’t ask this, but some might, and it’s important to include it in your system’s dialog,” Tettelin says. “Definitely think about the Edge Cases when you’re scripting conversations.”

5. Test these conversations in-house.

Feed your dialog flows into your working voicebot platform, then start having conversations. Record them. Note any unnatural pauses or missing information as you internally test your voice UI prototype.

Tips for Initial VUI Dialog Testing

When listening to your test recordings, look for moments where the conversation breaks down. “Maybe there are unnatural pauses, or you realize you need to add a bit more information to a specific step,” Tettelin says.

Then bring in help from across your organization.

“I like to make a little focus group that includes a person from each target audience,” Tettelin says. “They tell me what they like, and what they find weird. We record these conversations and listen back to see where there’s a bit of friction, where we could make things easier for the user.”

6. Publish a live beta version of your voicebot.

Once you’re satisfied with internal testing, it’s time to bring your creation to the world. But publication isn’t the end; it’s just the beginning. Monitor these first real-world conversations closely. See how people phrase their questions or commands, then adjust the bot’s responses to continually improve performance.

Tips for Public VUI Testing

“At this point, your voice assistant starts talking to real users,” Tettelin says. “That’s a very interesting part of the bot-building process, because a lot of people think, ‘It’s live, I’m done.’ But you have to really monitor those first conversations and see what people are actually asking the bots. How are they phrasing certain things? You can use those insights to optimize your bots and make them even smarter and even better.”

In order to complete all these steps, of course, you need very specific tools. Before you start building a voicebot prototype, here’s what you’ll need.

3 Essential Voice Prototyping Tools

With these VUI prototyping tools in hand, you’ll be ready to start building a voicebot in no time.

1. A flowchart-creation app.

Flowcharts help to keep your conversation scripts organized—and tools like Miro and Whimsical make it easy to build flowcharts. Use them as you design conversation dialog in Step 4 of our prototype-building process.

2. A self-service conversational AI platform.

To build a website, you need a publishing platform like WordPress. To build an AI voicebot, a conversational AI platform like Chatlayer fills the same role. This no-code SaaS solution provides everything you need to prototype, test, and deploy chatbots and voicebots on your own.

Chatlayer also provides an analytics dashboard, which simplifies the final stage of the prototyping process outlined above. Once you open your voicebot to real-world users, Chatlayer analytics provides detailed information about these interactions. You can identify the most common use cases, which languages you’re missing, and much more at a glimpse.

3. The right text-to-speech (TTS) voice.

A full suite of ReadSpeaker synthetic voices is natively available in Chatlayer. But for a more powerful brand experience, work with ReadSpeaker to build a custom TTS voice, designed for your organization alone. Custom-branded TTS voices provide a consistent and instantly recognizable identity, from the voice UI prototyping phase to day-to-day interactions across voice channels.