Voice assistants as specialists

A key to the success of voice assistants such as Erica and Pandora’s Voice Mode is their customization and singularity of purpose. These speciality voice assistants are built with speciality domains  to reflect more accurately how we seek advice and information from humans.

For example, when you ask your physician if an activity is too “taxing” given your condition, you wouldn’t expect her to refer you to an income tax software company. She would understand your meaning within the context of your visit. Similarly, if you ask your microwave to thaw two pounds of frozen chicken, the domain specific voice user interface will likely understand that command and not confuse the word “thaw” with “Thor” and try to find a movie on your TV, as a generalized voice assistant might. 

As customer demands for better voice user interfaces push more companies to provide meaningful and accurate experiences, we will stop expecting one voice assistant to act as the conduit for all other devices or to know everything. Instead, individual products and devices, customer service centers, and mobile apps will each contain the specialized knowledge needed to better serve niche customer use cases.

To take it one step further, custom voice assistants of the future can have personalities to match their function, such as a refrigerator that jokes, “Pasta again?” Giving these assistants distinct personalities is a natural progression for voice specialists with specific knowledge and expertise. 

Anthropomorphism—our innate human tendency to attribute human trains, emotions, or intentions to non-human entities—has already given rise to naming voice assistants and creating chit chat domains. Further humanizing them will be key to brand differentiation and product popularity as voice becomes our primary method of interacting with the world around us.

Voice assistants designed with the user experience in mind

According to the 2020 Adobe Voice Survey, 57% of voice assistant users surveyed are looking for greater accuracy. Presently, one in four requests is currently met with friction and frustration. While a variety of factors may contribute to inaccurate results, one of the greatest influencers is context and understanding intent. 

For voice assistants, meaning is derived from the content domains they access for answers. Content domains are like skills or abilities for Alexa and Google with one major difference: Content domains are areas of knowledge developed by subject matter experts and made available on voice AI platforms. While anyone can build a skill, quality domains are developed in a way that’s well thought-out by people who are experts in voice AI. 

When the voice assistant has access to every skill that anyone has built, the assistant loses the context of the conversation, unable to determine whether to pull from a cooking domain or one built for checking the weather. If your voice user interface is based on the Alexa or Google platform, you don’t get to pick and choose which knowledge base is more important and which are irrelevant to your use cases, you get them all. The result? A voice assistant that doesn’t know if she’s at the doctor’s office or with the tax attorney.

Specialized custom voice assistants built with expertly developed content domains can deliver the knowledge and accuracy lacking in the most popular voice assistants currently in use. Afterall, your refrigerator customers may like to get recipes, but most likely have no need for parking information. Benefits of developing a customized voice assistant based on an independent platform with an extensive library of content domains, includes:

• Developers can find the domains most relevant to their product
• Voice assistants provide the most accurate answers to unique use cases 
• The VUI will contain only those domains that add value to your hardware, service, or app
• You gain ability to create a voice assistant with in-depth knowledge in your subject area
• Companies can eliminate domains that don’t make sense for their users

Compound, complex queries use data from more than one domain

When we talk to another human, our neurons make connections in our brains to form links from one part of the conversation to another, allowing us to carry on a dialogue. For voice assistants, these connections can be made via a technology that links one domain to another, allowing the user to speak conversationally.

For instance, instead of saying, “What is the travel time to San Francisco airport from 5400 Betsy Ross Drive, Santa Clara?” and following up with “Is United Flight 794 departing on time?” voice AI technologies, such as Query Glue, allow the user to ask for things from different domain data sets, such as “How long will it take to get to SFO right now and is United Flight 794 departing on time?”

The same way a human brain can connect one thought or part of a conversation to stored information, voice assistants can retain information from one query to the next and combine domain knowledge using advanced AI technology and data sets that are connected one to the other. Technologies, like Deep Meaning Understanding® , allow users to ask follow-up questions and get responses based on their intent, extracting the meaning from a limited number of custom domains that understand the context of the conversation.

According to the recent Adobe survey, users of voice assistants are looking for more specialized applications, requiring voice assistants with specific domain knowledge. The most desired voice applications include:

• Getting travel directions – 39%
• Reviewing bank balances – 37%
• Updating work tasks and calendar events – 34%
• Making payments – 33%
• Making a restaurant reservation – 32%
• Booking medical appointments – 29%

Delivering accurate results for these applications will require a move from “know-it-all” voice assistants to those that can quickly respond with specific domain knowledge. Companies in the travel, banking, business, hospitality, and medical industries will need custom cross-channel voice solutions that unify the customer experience, while saving time and streamlining costs. These voice assistants must be equipped with specific knowledge and provide responses at a rate that far exceeds the current 75% accuracy now available on large voice assistant platforms.

Most of the desired applications identified by current users will also require an elevated level of user security, as they can include sensitive information. Privacy concerns will need to be addressed for users who don’t mind if the refrigerator knows they’ve had an extra donut and are happy to discuss a pulled hamstring with a virtual doctor, but don’t want Amazon to have access to that level of personal information.

Custom voice assistants provide a level of compartmentalization of not only knowledge, but also information storage and data retrieval—key to the ultimate success and adoption of voice AI.

Custom commands and domains for voice assistants

The best voice assistants are developed based on user knowledge collected and assembled from actual research. Knowing your customer, their expectations, needs, and desires is the basis of creating exceptional voice experiences. Custom commands and domains can augment available off-the-shelf domains to increase the functionality of your product or improve the potency of your customer service center assistant.

In some instances, the default commands associated with the existing library of domains match your user needs. In other cases, you’ll want to fine-tune those experiences with custom commands that match your users’ likely phrasing or the phrasing necessary to unlock the functionality of your product. 

In combination with domains that need no customization, such as the weather, or navigation, custom domains can be developed to address specific blocks of knowledge required by the voice assistant to address the specialized knowledge required to address unique user needs. 

Of course, you’ll want these domains to be developed by experts in voice AI and not rely on the “skills” type of domain creation that is available to anyone with a computer and internet access. Creating a conversational interface also requires that custom domains remain proprietary to your brand but still connected to the other domains available to your custom voice assistant.

Make onboarding and education easier with specialized voice assistants

One of the greatest challenges to elevating voice assistants as the first choice for user interfaces is education and onboarding. Frustrated by inaccurate answers, users often default to the simplest commands and revert to other methods of interaction to take care of specific needs. 

According to the Adobe study, an average of 63% of users sometimes don’t know where to begin, including those who use voice daily. The percentage is even higher (72%) among those who use voice less frequently.

Custom voice assistants with specialized content knowledge help address the challenges of onboarding and adoption by narrowly defining the use cases available. Think of it as sitting down at a restaurant with a six page menu filled with appetizers, pages of entrees based on pasta, chicken, beef, and vegetarian options, soups and salads, and a variety of sides and desserts. The choices become so overwhelming, making a decision is nearly impossible.