Virtual Assistants: How Perception is Affecting User Experience

With respect to people’s perception of Intelligent Virtual Assistants (IVAs), the age old statement, “You can’t judge a book by its cover” has never resonated more strongly with me than it does today. To further break this down, it helps to have an understanding of the types of IVAs out there. There are public IVAs that are typically found on mobile devices, like Siri and Google Now, and then there are enterprise IVAs, used by companies like United Airlines, Alaska Airlines, Verizon and the U.S. Army. According to a recent report by Forrester, only 55% of people who have interacted with an IVA in the last 12 months reported having a great experience. But before we go screaming to the hills that the technology is dead, I think it’s important to understand some of the reasons why people might have gotten a sour taste from their recent virtual assistant interactions.

High user expectations 

Because IVAs offer users endless possibilities, they oftentimes fail to contain user expectations. Take Siri for instance: When Apple branded Siri, they painted a nebulous picture of what a user could and couldn’t use it for. Apple customers’ expectations were influenced by enticing commercials of people on their morning jog, in their kitchen and riding in a taxi – interacting with their phone’s IVA (Siri) in very vague terms via voice commands. Intriguing, yes – misleading, I think so. There were no limits or boundaries established for its capabilities. As such, Siri set itself up for not only fantastic experiences – but poor ones as well.

Public IVAs solve a broad range of problems and can address simple phone tasks as well. Siri, for example, lets you use your voice to send text messages, call contacts and schedule calendar events. Public IVAs are extremely useful for performing tasks like these that are within their scope. Ask a question about sports or weather, and the system still performs reasonably well, delighting users with an answer and possibly even a joke. But take it one step further and ask Siri about booking a flight or checking your bank balance, and it submits a web search – which does nothing to fulfill the request. As the user progresses farther away from phone-based actions and towards more personalized questions, they’re not only moving beyond the bounds of public-IVA technology, they’re also forming an opinion – most likely a lowered expectation – of what a “personal, virtual assistant” is capable of. When dealing with a public IVA like Siri, chances are good that, because of their lowered expectations, users will write-off a less than helpful experience as worthy of a good laugh.

User expect more from enterprise IVAs

When consumers interact with enterprise IVAs, they are coming to a site or mobile application for a very specific reason, usually involving something personal like a question about the balance on their account. Because of this mindset, enterprises must build-out deep language models that know as much as possible about the enterprise’s domain and that have access to data from backend systems. This is the only way to ensure that an IVA interaction will be meaningful and personalized to each user’s unique request. As an example, when you go to AlaskaAir.com, you are visiting to accomplish a specific task. If you want to book a flight and you launch Jenn, their IVA, you can simply say, “I want to fly from Seattle to LA on Friday and return Sunday.” Jenn’s deep knowledge of Alaska’s systems allows her to identify the intent of the question, mine and normalize the appropriate data and submit it to Alaska’s booking engine to show the proper results.

As you can imagine, a person interacting with Jenn counts on receiving accurate, useful information. As a result, when they get the appropriate answer, they expected it, and if they don’t, they are severely disappointed and may, in fact, call the enterprise – which adds even more frustration to their overall experience.

At Next IT, we take great pride in our IVAs and have honed our implementation skills for over 11 years. This invaluable experience gives us a unique perspective on how to properly build IVAs that will yield value instead of coming across as a gimmick that further annoys users.

Let me know how you view experiences with the IVAs out there today.