Voice AI Design
Oct 05, 2021
7 MIN READ

5 Tips to Improve Voice Assistant Design and Customer Experience

Better voice user interface designs deliver improved user experiences essential for creating loyal customers and brand evangelists. According to Statista, in 2020, there were 4.2 billion voice assistants worldwide, which is expected to grow to 8.4 billion by 2024. The exponential growth of voice assistants will make getting your brand heard amidst the vast sea of voice assistants even more difficult. Soon, brands will need differentiated experiences that speak to their users’ needs, expectations, and desires to stand out and capture customers’ and prospects’ attention and remain competitive.

As manufacturers already know, user experiences that cause frustration lead to brand and product abandonment. Given their relative newness to the market, frictionless voice experiences are even more critical for brands hoping to improve user experiences. Because first impressions are often lasting ones, companies deploying voice assistants will want to make sure their voice user experience delivers on the promise of a convenient, easy to use, accurate, and responsive hands-free experience the first time it’s used. Following voice AI design best practices, brands can create voice assistants that address user needs and deliver a competitive edge from the first voice interaction.

Many aspects of good voice AI design originate in knowing your users—what their needs, wants, and expectations are, as well as their geographical location, age range, accents, culture, and commonly spoken languages. Brands also need to understand the environment where the voice assistant will be used, what content knowledge customers expect from it, and what user needs and cultural elements the voice assistant should understand. Having knowledge of these elements builds the foundation for a voice AI strategy that will inform other decisions along the way.

Whether you’re just getting started on your voice AI journey or you’re looking for ways to improve your current voice assistant, keep these 5 elements of voice AI design in mind: 

  1. Noisy environments
  2. Specialty domains
  3. Multilingual
  4. Cultural bias
  5. User testing

Accommodate noisy environments

A voice assistant must, above all else, deliver on its most important and basic function—to understand the user. Regardless of the use case, background noise of some ilk is always a consideration in voice user interface (VUI) design. Whether it’s road-noise, background music, or just other voices, voice assistants must be able to differentiate between the user and the background. While voice AI technology can help with features such as echo cancellation and beamforming technologies, voice assistant design best practices require developers to consider a variety of conditions to improve accuracy.

Here are 6 keys for designing voice assistants for noisy environments: 

  1. Know your user’s environment 
  2. Choose the right microphone 
  3. Choose linear noise reduction components
  4. Add zone control with source separation 
  5. Include more data in training models
  6. Allow for interruptions and barge-ins

One of the most important considerations when designing for noisy environments is knowing what type of background noise will need to be introduced into training models. For in-car voice assistants, that may be other passengers and road noise. For mobile apps, that could be street noise or background conversations. For smart home technology, the user could have music or the TV playing in the background. In some cases, the voice assistant will simply need to differentiate between the user’s voice and the voices of other people in the room.

One of the most important considerations when designing for noisy environments is knowing what type of background noise will need to be introduced into training models.

In addition to following VUI best practices for design, you’ll also want to choose the right microphone and linear noise reduction components. The right microphone with good directivity towards the speaker can ensure that reverberation coming from other angles towards the microphone will be lessened.

There are also strategies that brands can put into play during the development of the voice assistant for noise cancellation. By training acoustic models with noise, you’re teaching them to recognize the human voice and disregard the background noise. When training models, you’ll want to introduce it to as many types of background noises as possible (e.g. engine noise, background conversations, road-noise, crying baby, music, television, etc.) to give the model as many scenarios to understand and adapt to. 

By following these tips for designing voice assistants in noisy environments, you’ll be able to increase your accuracy rate and customer satisfaction. A voice assistant that can understand users wherever they may be is a great way to differentiate your voice assistant and create positive user experiences.  

Smart voice assistants with specialty knowledge

Voice assistants that are designed for specific purposes, such as driving, operating in-home appliances, watching television, listening to music, and other special use cases, need specific knowledge to complete the tasks that users expect of them. Custom voice assistants that are equipped with custom content domains can lead to greater customer satisfaction and adoption rates by providing a superior user experience. 

Custom voice assistants that are equipped with custom content domains can lead to greater customer satisfaction and adoption rates by providing a superior user experience. 

Having specialty domains for voice assistants is essential for improved accuracy. Equipping your voice assistant with the right information begins with conducting a vast amount of user research to determine user expectations, needs, and desires. In some cases, default commands in a library of domains may serve your users well. In others, such as an in-car voice assistant, specific domains that specialize in navigation, entertainment, cabin controls, information, and communication are required. In other cases, like a smart oven, users will want it to know various temperatures for cooking but will likely not expect it to know the temperature outside. 

With a voice assistant programmed to know a vast amount of information and perform complex tasks, so many users still only ask voice assistants for the most basic commands, such as setting timers and playing music. That’s where onboarding and education come into play. By informing your users through tips, reminders, and suggestions, you can educate them on how to access your specialty domains and receive an exceptional user experience. 

Whether you’re just beginning your voice AI journey or evaluating how to improve your current voice assistant, it’s important to determine what domains will best serve the function of your voice assistant and your user’s needs. 

Multilingual voice assistants speak to your users

Having a multilingual voice assistant is essential for widening your target audiences and geographic locations. Users feel more comfortable speaking in their native language and will use your voice assistant more if they have natural interactions with it. 

Designing a multilingual voice assistant isn’t as simple as translating commands into different languages. Elements, such as localization—including targeted content and formats—are vital to ensure the voice assistant responds in a culturally acceptable manner and can understand speech in the context of different rules of grammar.  

Users feel more comfortable speaking in their native language and will use your voice assistant more if they have natural interactions with it. 

Here are 5 key steps for creating a multilingual voice assistant:

  1. Make critical voice strategy decisions 
  2. Figure out what’s important to your brand and customers
  3. Follow proven best practices for training your voice assistant
  4. Avoid common misconceptions and missteps in VUI design 
  5. Find the right voice AI platform partner

Before beginning to design a multilingual voice assistant, developers need to make some important decisions, such as how many languages the voice assistant should understand fluently, which geographic locations should be targeted, what competitors are doing, and how the voice assistant will be used. Knowing answers to these questions will help map out how the voice assistant should be trained. 

When beginning to train your voice assistant to understand multiple languages, it’s vital to follow best practices and avoid common misconceptions in VUI design, such as conflicting data sources, overfitting, and relying too much on intuition. 

Brands may consider finding a voice AI partner who is an expert in designing multilingual voice assistants. The technology and capabilities can take years to develop, and buying a voice AI platform that is already capable of speaking the languages you are targeting can save a lot of time and money. 

Designing to eliminate cultural bias

When starting out on your voice AI journey or re-evaluating your current strategy, it’s important to ask whether your voice assistant safeguards against racial, gender, accent, age, and regional bias. Overcoming cultural bias in voice AI design is essential to creating a voice assistant that speaks to all users and communicates brand values of inclusivity and acceptance. 

When starting out on your voice AI journey or re-evaluating your current strategy, it’s important to ask whether your voice assistant safeguards against racial, gender, accent, age, and regional bias.

A possible solution for overcoming racial, accent, and regional bias is to ensure that your voice assistant can understand multiple accents, especially from your targeted users and locations. While most brands will think of the difference between a West coast, East coast, Spanish, or Chinese accent, other accents such as African-American Vernacular need to be considered just as much as the rest. Otherwise, the voice assistant is essentially excluding that group of users and contributing to discrimination. 

Gender bias is a widespread concern for the voice AI industry, with a majority of voice assistants using a female voice over a male’s. Possible solutions to overcoming gender bias include having a non-binary voice assistant, having more women on the design team, and ensuring a female voice isn’t chosen because of stereotypical submissive or nurturing qualities. 

With the increasing rates that voice assistants are adopted, it’s essential to make cultural biases a vital part of your voice AI design. More users will feel comfortable using your voice assistant if it is free of biases, and your brand will also be paving the way for an inclusive future. 

The critical need for user testing

User insights can be critical information at many points in voice AI inception, design, and deployment. It’s essential to perform user testing for voice assistants before you begin the development phase, so you have all the key pieces of information on your users and their wants before you begin developing—saving you time, money, and effort. 

It’s essential to perform user testing for voice assistants before you begin the development phase, so you have all the key pieces of information on your users and their wants before you begin developing—saving you time, money, and effort. 

While user testing is important to start before the development phase, it’s also vital to conduct it during design and after deployment as well. Many new elements and questions may arise during the design phase that can benefit from user testing. User testing is especially important after deployment since the voice AI process should never be a one-and-done strategy but be continuously tested and improved upon based on new technological developments and further user feedback. 

Here are 5 types of user research that every brand developing a voice assistant should consider: 

  • Focus groups
  • In-lab testing
  • Remote testing
  • Customer surveys
  • AB testing

When conducting user research, you’ll want to make sure that you test your actual user population to get a clear and accurate understanding of who your target audience is and what their needs are. It’s also important to have regular communication about the testing and results with all the key stakeholders. For the best research outcomes, it’s better to over-communicate and make sure everyone is on the same page and has the same goals than risk miscommunication.

Additional VUI design tips

Interested in learning more about voice AI design best practices? Here are some more blogs you can check out for a deeper dive:

At SoundHound Inc., we have all the tools and expertise needed to create custom voice assistants and a consistent brand voice. Explore Houndify’s independent voice AI platform at Houndify.com and register for a free account. Want to learn more? Talk to us about how we can help bring your voice strategy to life.

Kristen is a content writer with a passion for storytelling and marketing. When she’s not writing, she’s hiking, reading, and spending time with her nieces and nephew.

Interested in Learning More?

Subscribe today to stay informed and get regular updates from SoundHound Inc.