Overcoming the Top 3 Challenges of Voice AI Adoption

Adding voice AI to your product roadmap might seem as daunting as planning a mission to Mars, but with this guide, it doesn't have to be. We’ll walk you through the three biggest challenges of adopting voice technology and provide expert advice on how to overcome them. Explore the world of voice assistants and land a voice-first strategy that drives adoption, delivers a competitive advantage, and grows market share.

The Launch: Make Your Voice Discoverable

Accelerate voice assistant adoption and create exceptional user experiences.

The evolution of user expectations and growth in market innovation have been the drivers behind the race to implement voice AI into everything from cars, to hotel rooms, to appliances, to mobile apps.

In this brave new world of voice assistants, a customer-first perspective, driven by data, has become the standard for brands seeking to implement a long-term voice strategy. As this new era unfolds, market leaders have recognized the need for ongoing user education and continued iterations to ensure higher user adoption and device discoverability.

We see consumers flocking to brands that complement their products with high-quality services and experiences. Voice can be a tool to not only sell products but to deliver experiential value that builds long-term loyalty.
Shanna Walia, strategy analyst, RAIN

In the early stages of voice exploration, organizations anxious to implement a voice-first foothold often defaulted to installing an established third-party voice interface into their products. Although this route was considered the simplest and least expensive method of implementing a voice assistant, it creates a disconnect between the brand and its customers and loss of control over user experiences and data. Early users of these products became distrustful as it became clear that their interactions with the product or device were resulting in targeted advertising and marketing efforts.

of business leaders plan to increase their investment in voice in the coming year.
Source: Adobe Survey

According to Dennis Yang, co-founder and CPO at Dashbot, “Launching your voice AI application is only the first step. You should expect to learn an immense amount from the usage of your application. Conversational applications provide a richness of understanding that websites and mobile applications can never have.”

Dennis Yang.
Co-Founder & CPO
Dashbot
By analyzing and understanding the conversations with your application, you will be able to understand where users are confused, where users are delighted, and the nuanced area in between.

As voice becomes more established, there is still a lack of uniform thinking about what a voice strategy needs. Increasingly, this confusion is giving way to the understanding that the one thing brands don’t want to do is hand over their customer relationships and data to third-party vendors—who may in fact be competitors—or who may be using that data for their own commercial purposes. The combination of the users’ needs for privacy and a company’s desire for brand loyalty have superseded the desire to rush into what originally seemed like the easiest and least expensive solution.

A new era in voice innovation began when brands recognized the need to retain customer connections and control the collection and use of the resulting data.

With those goals in mind, the imperative is clear: Deliver a voice interface that is an extension of your brand and is personalized to users based on your direct relationship with them.

The race to innovate has begun in earnest and brands in every industry are looking to voice AI as a way to differentiate their products, services, and devices from the competitors. Determining the limits of what your voice assistant can deliver begins with listening to your customers and testing your systems thoroughly before going to market with a solution.

Invocations are the new URLs
Peter Stewart, author/consultant/speaker, The Smart Speakers Podcast
Best Practices for Launching a Voice AI Strategy
1

Start a conversation with your customers

Talking to your customers is different from actually having a conversation. Conversations require careful listening and appropriate responses. As you determine your voice-first strategy, be sure to put your customers at the center of the conversation. Instead of designing a device that answers all the questions anyone would ask in any environment, limit your dialogues to those conversations most critical to your users in the unique environment where they will be using your product or device.

At every step of development, make championing the user experience your goal. Consider the user as you write the dialogue, determine which domains to include and design for personalization. While integrating voice into a product is an exciting adventure, overreaching and over-designing may actually lead to user disappointment.

Begin with a simple, focused mission to answer your user’s most likely queries in the most responsive and accurate way possible. Once you’ve successfully delighted your consumers with a product that delivers on its promise to improve their experiences, continue to iterate organically based on customer use cases and data.

According to Michael Hoesten, director of client engagement for Skilled Creative, “Before diving into scripting, it’s important to take the time to explore consumer behaviors, possible use cases, and your user flow map. If your user flow isn’t sufficiently developed, you’ll find yourself backtracking in scripting.”

For instance, if you’re designing a VUI for an automotive application, you’ll want to ensure that your users have the best navigation, car control, and local search domains and dialogues. In the first iteration of your product, it’s critical that the core use cases are present and work well before other functions, such as sports news, the day’s stock numbers, and astrological forecasts are available. However, if your data shows that your users are consistently requesting these types of information through your query data, consider adding them into a future iteration of your product interface.

Voice is a prime example of the old adage. Sometimes you need to slow down to speed up.
Michael Hoesten, director of client engagement, Skilled Creative

While your voice assistant should always be ready to help, resist the urge to be too helpful. Personalizing notifications and suggestions can quickly move from being perceived as beneficial to becoming an intrusive annoyance.

James Poulter, CEO of Vixen Labs notes, “Many voice experiences fail at the first hurdle because they fail the 'long menu' problem. If you go to a restaurant and there are 100 things on the menu, it’s so much harder to choose what you want, than with a limited but carefully chosen list of options. Remember that users have chosen voice to get to the point, not be dragged through menus, drop downs, and filters of a website.”

He adds, “Design with a conversation in mind, not a menu system. Make it easy for a user to get to what they want quickly through structuring suggestions for them rather than asking them to question every detail.”

In the voice AI world, those who simply install a third-party voice assistant into their products and devices are not likely to create lasting customer connections. The missing link between third-party voice assistants and custom VUIs are the custom domains that enhance the user experience within a specific product.

Michael Hoesten, director of client engagement at Skilled Creative adds, “We should begin by thinking in terms of how voice can function as a control layer on top of all our digital platforms.”

In practical terms, if you are designing a speaker to be in used in a hospitality setting, in addition to providing weather and local search information, you’ll want to be sure your guests can control the room environment and get informationabout that specific hotel or cruise ship. For example, we recently partnered with HARMAN’s hospitality cloud AI to bring the power of voice to the hospitality industry. The solution provides sophisticated, voice-enabled “smart room” experiences for guests in hotels and onboard cruise ships.

Beating the competition and creating stellar user experiences require that the customer remain the focus and that the product or device provide something more than what the consumer can get via tapping and swiping a screen. Keeping the customer at the center of the conversation is key to a voice-first approach.

Kane Simms
Co-founder & CEO
VUX World
Without a solid use case and a truly engaging user experience, generating revenue will be extremely difficult.
2

Pre-launch data

Before the countdown begins, plan a few test sequences to discover deficiencies and improve the user experience. Taking the time to iterate your processes and test your voice interface experience with a small group of users will pay dividends once the actual product is launched.

Plan and communicate the various types of testing and the expected outcomes to company stakeholders. Emphasize the need for well-designed tests and product iterations based on the results of voice trials. Include a robust variety of testing methods and testing subjects, including:

DTT
Digital Traversal Testing by developers and QA testers
VRT
VUI Review Testing by VUI designers and speech specialists
WOZ
Wizard of Oz Testing by prospective users

Skipping any one of these critical testing phases may result in a VUI that goes to market and performs below expectations. Each testing phase should be followed by an iteration phase to correct any issues or errors discovered during testing.

When you're first starting out—don't forget to tap into your existing sources for training data. Customer service chat logs or IVR transcripts are great places to look for utterances.
Dennis Yang, co-founder & CPO, Dashbot

Before you expose the product to users for testing, take the time to conduct some tests internally with developers, designers, and speech specialists. DTT and VRT tests with internal teams give you the ability to test for, and fix, known or expected issues. Put testers in the user environment to allow them to interact with the design from the user perspective.

Once internal testing is done, WOZ testing provides an opportunity to discover something you don’t already know about the user interface. In this scenario, test participants interact with what they perceive as a live VUI, which in reality is being operated by a researcher in another room (behind the curtain). During these tests, set your goals to discover specific elements of the voice interface, including:

Search intent
Error recovery performance
Language and terminology usage
Perception of the VUI persona

It’s important to keep the user and the user’s environment in mind when conducting these tests. Taking the VUI out of context may lead the team down pathways that are not consistent with the purpose of the voice assistant. Focus groups, usability studies, user surveys, and interviews are additional venues for understanding user intents and expectations for your voice application.

Kane Simms, co-founder & CEO of VUX World advises, “The more often you can put your prototype in front of real people, the more you’ll learn what people expect from it, how people talk to it, and what capabilities you need to provide.”

Lastly, brands must set clear goals for teams to rally around. What part of the customer journey is voice meant to impact? And what is considered successful engagement? Limiting the scope of the VUI may actually result in higher user engagement and satisfaction.

Brandon S. Kaplan
Founder & CEO
Skilled Creative
Voice requires an iterative approach. Give yourself the luxury of an iterative strategy after launch, and the result will be a dramatically enhanced experience with continually improved performance.
3

Educate and engage your audience at every step

Before your voice experience is ready to go live and encounter your users in the real world, have a plan to ensure your users are aware of the functionality available to them. The first steps in this process will require marketing to lead the way and get the word out that a voice interface is available to improve your product experience. If you’ve engaged your marketing teams early on, they will be helping champion the voice initiatives internally as well as devising launch plans for once the interface is ready to be pushed live.

When building your launch plan, you may not want to go quite to the extremes that Mercedes-Benz did in their 2019 Super Bowl ads, but you’ll want to make sure your audiences are made aware of your innovations.

Dave Kemp, creator and publisher of FuturEar adds, “It won’t matter how awesome the voice experience is, if customers are not aware that it exists. Therefore, companies should lean on their legacy marketing and communication channels to help grow the awareness of the voice experience.”

Simply having a voice assistant embedded in your product will not be enough to ensure its success. Although some users will embrace the new interface and know instinctively how to use it, others may need more prompting and information before they are comfortable talking to a voice assistant.

When designing user onboarding and ongoing education experiences, keep these three goals in mind:
  • Stimulate exploration of functions
  • Elicit adoption
  • Encourage expanding usage

User education begins with the marketing efforts that commence when the product is introduced into the market, but it doesn’t stop there. Once a purchase is made, there may be an opportunity for on-the-spot education. A salesman at a car dealership can provide a walkthrough of the basic and most popular voice queries for drivers. The hotel front desk agent can provide some basic information about the in-room smart speaker, including privacy information and special features—such as the ability to ask for all data to be deleted upon check-out.

Written materials such as a laminated card in the hotel room or a short user’s manual in the car’s glove box will give users something to refer to beyond a quick walkthrough—or when there isn’t time for personalized onboarding. A short and visual user’s guide can also be included to get consumers started with the product.

High user churn is a problem for voice applications. To improve retention, design voice apps that educate users on how to get the most out of them from day one, and let them personalise the experience.
Carl Robinson, Host, Voice of Tech Podcast

If you haven’t planned for ongoing user education, you’re likely to see a loss of traction after the first few interactions. Users will either become frustrated or bored if they aren’t able to easily discover new functions and ways to get more value from the voice assistant. New features, functions, and tips can be announced via existing communications channels, such as email, text, and voice messages.

When and how these communications are delivered must be determined based on the user’s need to know, and a limited cadence of frequency, to avoid being perceived as overbearing and disruptive.

Dave Young
Lead Creative Technologist
Skilled Creative
Don’t be self indulgent! There is a delicate balance between a great creative experience, and too dense of a conversational script.

Voice assistants can help expand a user’s depth of experience with the product through some interactive dialogues along the way. Deploying prompts strategically during an interaction with the voice interface can lead users to discover greater functionality and value in the moment. Learning how to get more from the voice assistant organically is more likely to lead to the user repeating the action in the future, versus simply reading about it out of context.

When designing helpful prompts, keep these best practices in mind:
  • Present options clearly and concisely
  • Avoid questions that can be answered with a yes/no response
  • Provide an option to get additional help
  • Deliver step-by-step information one step at a time
  • Group information into memorable chunks of three or four items
  • Ask only necessary questions
  • Avoid too many confirmations

No matter how your user education is delivered, avoid the impulse to let your customers know everything they can do up front. Follow general teaching best practices guidelines and coach users a little bit at a time. Ideally, the user will have a chance to practice using a new query or a new function before receiving the next tip or prompt.

Hannah Paxton
Strategy Analyst
Rain
It’s important to continually update your voice experience based on how users are behaving with it. An informed feedback loop will increase the retention value of your experience.
4

Using post-launch data

Once your product has been released, your user data will be your most important method of gaining feedback on the efficacy of your voice assistant. When a third-party voice interface is used, brands lose this essential source of product usability and engagement. Deploying a voice assistant in your product or device, without the ability to own your data, is akin to launching a rocket without a guidance system.

According to Hannah Paxton, strategy analyst at RAIN, “Metrics can only mean so much if they are not being used through the lens of defined KPIs. Success with voice could look very different from other marketing activities, and it’s important to set expectations, know what data patterns you’re looking for, and be ready to iterate your findings.”

Without owning your data, you’ll never know if your rocket ship reached its destination or what it discovered once it arrived. Partnering with an established and reliable voice AI platform provider to develop a customized voice user interface, with a branded wake word, puts you back in the captain’s seat of your own product journey.

It’s important that any practitioner of the technology understand the layers of the channel, including NLP, NLU, TTS, and the 3rd party players who allow you to build your own enterprise interface.
Dave Young, lead creative technologist, Skilled Creative

When you have control over your user’s data you accomplish two important goals. First, you can provide greater privacy for your customers and ensure that their interactions don’t result in annoying pop-up ads on every device they own. Secondly, you can plan for product iterations based on how your customers are using your product or device. At this point, you’ll want to start monitoring interactions to understand what’s working and what needs adjustment. Take the opportunity to promote voice use cases that you’ve observed being used successfully as a way to help get other customers on board.

The data and analytics you gather from actual user interactions should measure the following:
  • Top intents (What the user is asking for, like “Find me a restaurant”)
  • Parameters (The specific contextual request, like “..within 5 miles serving Chinese”)
  • Sentiment analysis (The language used, as well as the tone or pitch of users)
  • Pathing (Includes the steps of the conversation as well as what actions users take afterward)
  • Errors and null statements (Demonstrate where devices and platforms fail their users)
  • Length of sessions (Indicates depth of user engagement)
  • Frequency of use (Identifies areas of additional user education)
James Poulter
CEO
Vixen Labs
The best way of getting a true gauge on what your customers want is by launching agile products that give you signals of what they are interested in and how and when they are asking.
5

Focus on ongoing optimization

By now, most people are aware that voice assistants can not only provide information and helpful interactions, they can also collect information about the user. Since this data is often used by the largest providers of voice interfaces for advertising and marketing purposes, people have become distrustful of devices.

Many of them have experienced at least one negative event or frustration with a voice-activated device. While there is currently no way to avoid all miscues, misdirects, and machine errors; planning for a regular cadence of VUI iterations will build the expectation of ongoing optimization into your roadmap. It will also help avoid making every adjustment a new change ticket for the voice team.

Be sure you are prioritizing the user’s goal before offering to unlock premium features and functionality such as multiple accounts, transactional tiers, messaging, and more.
Heidi Culbertson, CEO, Marvee

Continued iterations of your voice assistant—based on the data you collect from your users—will ensure better experiences and reduce the likelihood of ongoing negative interactions on social media and low-scoring product reviews.

As part of that experience, be transparent with your users about the data you will be collecting. Let them know that their personal security is not at risk and “listening in” will not result in putting a target on them for advertising. Furthermore, communicate that their data will be anonymized and not traceable to them personally.

When surveyed, people admitted that they would like their devices to collect data, if that data were used to improve the functionality of the device and their experiences with it.

However, people do not want their privacy or security compromised in order to make a better product. Be transparent to build trust with your customers. Then, make good on that promise and continue to improve the product based on the data and analytics you collect.

Consider the four F’s when developing a successful Skill: It needs to make a task FASTER; minimise FRICTION; deliver content that’s FRESH; and have a name that’s intuitive, memorable, and FRONT OF MIND.
Peter Stewart, author/consultant/speaker, The Smart Speakers Podcast

As you prioritize iterations of your VUI, put error recovery at the top of the priority list. While humans naturally correct for communication errors, voice assistants need to be programmed to adjust when there has been a breakdown in human-to-machine understanding.

When the product responds incorrectly, can’t perform the function requested, or simply doesn’t reply, users get frustrated and may give up on the experience. Continually iterate your voice assistant to keep the conversation going between your product or device and your users. Make the most of those interactions with voice interface experiences low in errors and high in engagement.

“Now is the time to create functionality that is useful for children, the elderly, and everyone. Not just Millennials,” adds Stewart.

More advice from the experts:
Understanding who the customer is—their stage of life, social conditions, generational vocabulary, where they already spend time, etc.—improves VUI design and better utilization of qualitative data to meet users where they already are. - Heidi Culbertson
We have so many options to engage with brands in today’s world. It is a missed opportunity not to take advantage of cross-channel exposure. Not everyone uses Alexa, or Google, or a host of other voice assistants on the market. Meet your customers where they choose to be. - Bhawini Naik
Voice applications are notoriously difficult to discover. Most users don’t even know they can install third-party voice apps. To maximize your investment in voice, you must market your app to create awareness. If you have multiple voice apps, consider implementing some cross-promotion between them. - Carl Robinson