Speech Technology Magazine

Siri-ous Influence: Enterprise Virtual Assistants Have Arrived

First came Apple's Siri for the iPhone. Now there's Nina, Lexee, and Sophia for business use.
By Judith Aquino - Posted Nov 10, 2012
Page1 of 1
Bookmark and Share

On the day a businesswoman is scheduled to fly out of town, she is faced with a sudden emergency that requires her to leave for the airport immediately. While rushing outside to hail a taxi, she realizes she needs to find and book a new flight. Pulling out her smartphone, she asks, "Can I get on an earlier flight to Boston?"

A reassuring automated voice responds, "Yes. There is space available on flight 1215, departing at 2:45."

"Perfect. Book it," the woman says as she climbs into the cab.

In this scenario, there's no searching or typing to find a flight. It's done in a few seconds by a virtual digital assistant that knows her schedule and transactions. It's not pie-in-the-sky technology. In fact, it's one of the scenarios that Nuance Communications depicts in a promotional video featuring its speech-enabled personal virtual assistant, Nina.

A year after Apple introduced its voice-activated virtual assistant, Siri, app developers are betting customers will benefit from voice-enabled virtual assistants designed for the enterprise.

Already, there's Nina and Lexee. Created by Nuance Communications and Angel, respectively, both are software development kit (SDK) solutions that add voice-enabled customer service capabilities to mobile applications. A bank, for example, could apply Nina or Lexee to its mobile app, giving customers a virtual assistant that will verbally assist them in checking their balance or other tasks related to the bank. There is also Sophia, which Taptera designed to serve as a mobile voice-enabled virtual assistant for Salesforce.com users.

Lexee, Nina, and Sophia—which were all launched a few months ago—arrived at a time when more than half of mobile subscribers use a smartphone. By July 2012, 55.5 percent of U.S. mobile subscribers owned one, up from 41 percent in July 2011, according to Nielsen. And while the number of smartphone owners is increasing, so are the tasks the devices are being used for, according to Gene Alvarez, research vice president at Gartner.

"When our flight has been cancelled, we'll use our phone to book another flight or use it for help finding our hotel—we want our phone to help us when things go wrong," remarked Alvarez at IBM's Smarter Commerce Global Summit in September.

Virtual assistants are a natural outgrowth of our expectation for on-demand service, wherever we are. By 2015, 50 percent of online customer self-service search activities will be done via a virtual assistant for at least 1,500 large enterprises, according to Gartner.

Part of the appeal of virtual assistants is their ability to "let people express their needs more naturally than having to navigate a visual interface or a conversation tree like 'For X, press 1,'" comments Matt Lease, assistant professor in the School of Information at the University of Texas at Austin. "Virtual assistants let people reuse the same communication skills they have already developed for communicating with other people," Lease states.

In addition, speaking to a virtual assistant on a mobile device is convenient in situations when people have difficulty using a small keyboard on a phone or are occupied and unable to read information on a screen.

A virtual assistant can also be a way for an organization to set its customer experience apart from that of other companies. More than two-thirds of customer experience leaders recently surveyed by Forrester Research said their firms want to differentiate their offerings based on customer experience. As numerous companies have shown, delivering great customer service can lead to decreased operational costs, increased purchases, and long-term loyalty.

Enterprising Assistance

This past August, Nuance Communications introduced its Nuance Interactive Natural Assistant (Nina), a voice-powered app that lets companies add speech-based virtual assistant capabilities to their iOS and Android mobile customer service apps.

Nina uses Nuance's speech recognition technology, combined with text-to-speech, voice biometrics, and natural language understanding technologies to "deliver an interactive user experience that not only understands what is said, but also can identify who is saying it," according to the company.

If you think Nina sounds a bit like Siri, you would not be mistaken. Siri gave voice-enabled virtual assistants "a huge stamp of approval," acknowledges Robert Gary, president and general manager of the mobile care solutions unit at Nuance.

While Siri has driven awareness of the power of speech in virtual assistants, Nina is different, Gary insists. "Nina is a virtual assistant for the enterprise," he says. "She has very different business objectives [from Siri] and is focused on delivering customer service in a specific enterprise space."

Nina consists of the following: the Nina Virtual Assistant Persona (includes customizable voices and visualizations); an open SDK; and the Nina Virtual Assistant Cloud, which lives on Nuance's servers and powers most of the service's features. Nina understands U.S., U.K., and Australian English, and Nuance plans to provide support for additional languages in the coming year.

Nina validates a user's identity by storing that person's unique voiceprint. A user can then access his or her personal information by uttering a programmed phrase. To protect against someone using a recorded clip, Nina includes a security measure called playback detection that identifies audio segments that do not match the authenticated segments. To make sure that the app is operating correctly, Nina also lets companies take a snapshot of real-time call metrics on a dashboard, run historical summary reports, and view detailed information for individual calls.

USAA, a financial services provider that serves 9.1 million customers, is the first company to test the virtual assistant within its mobile app. At the time this article was written, USAA Bank had begun testing its voice-enabled app among a group of customers.

Customers who have been given access to the app can ask questions like "How much did I spend last week?" or "What's my checking account balance?" They can also schedule payments with prompts like "Pay my USAA credit card on Friday." The feedback from customers has been positive, according to Nicole Alley, a senior communications partner at USAA. "Since we launched the test last month, we've found customers are already using the app to answer questions about their accounts, such as their account balance, and to pay bills," Alley says.

Although the enhanced app is still in the beta phase, USAA customers are likely to welcome an app that makes it easier to use the bank's services, says Alley, who notes that even without the voice-recognition feature, the USAA mobile app has been downloaded more than 4.5 million times.

"Time will tell how many customers use the app," Alley notes, "but anything that helps our customers resonates with them, and we hope that the virtual mobile assistant will allow us to make their lives easier."

The app is expected to be launched to all USAA members in 2013.

Smart and Simple

Soon after Nuance launched Nina, Angel, an interactive voice response provider, unveiled its own mobile SDK for iOS and Android apps at this magazine's annual SpeechTEK conference.

Dubbed Lexee (based on the word lexicon), the SDK lets developers add speech-based virtual assistant capabilities to mobile apps. Lexee is supported by Angel's CX platform, which includes business intelligence analytics, voice biometrics, and SMS capabilities.

What sets Lexee apart from other voice-enabling virtual assistant apps is its ease of use and the analytical data that it provides developers, maintains Kelly Weinhold, product strategist at Angel.

"Even users without any technical acumen or coding experience can create their own voice flow with Lexee," Weinhold says.

Lexee uses a point-and-click system called SiteBuilder that allows nontechnical users to apply voice to any customer interaction. A subsidiary of customer engagement provider MicroStrategy, Angel also applied MicroStrategy's business intelligence software to Lexee.

"When enterprises create voice-enabled assistants for their customers, it's critical that they know how well they're working," Weinhold notes. "We provide the analytics that will show companies how many customers are using the virtual assistant, if [the virtual assistant] is successfully completing the task, and if not, where are they leaving the transaction."

As an example of what companies could do with Lexee, Angel added voice-enabled features to a Salesforce.com application. Instead of thumbing through data in their Salesforce.com application, sales managers and sales executives can locate the information they need by voicing their questions to the app, explains Jonathan Hotz, Angel's senior product manager.

"As a sales manager, I could ask Lexee, 'What are John McClane's latest deals?" Maybe John did really well that quarter and I want to congratulate him. I can then ask Lexee to call John's phone number," Hotz says. "I've got that interaction between the Lexee app and the phone, making it a multichannel environment."

Corporations from various verticals such as financial services and travel have expressed interest in deploying Lexee for their customers, according to Hotz, who declined to name the companies.

Lexee currently understands U.S., U.K., and Australian English. Due to interest among Spanish- and Japanese-speaking companies, Angel plans to add these languages to Lexee, in addition to others, in the coming months, Hotz says.

Salesforce.com Support

Taptera, a mobile business applications developer, debuted its voice-to-CRM iOS virtual assistant app, Sophia, in September. Using natural voice understanding technology, Sophia lets Salesforce.com users update their sales entries with call details and other data from their mobile device.

"We like to think of Sophia as a child of Siri and Salesforce," comments Chris O'Connor, cofounder and chief executive officer of Taptera. "One with a deep integration of contextual voice technology and understanding…in an intuitive mobile application."

Sophia was designed with the busy sales executive in mind, O'Connor explains. After logging on to the app, which runs on Salesforce's platform, an executive can tell Sophia that he or she would like to update an account and dictate the information to Sophia.

Taptera estimates that dictating notes and updating entries with Sophia saves users 15 minutes per contact. In case a user wants to type the information, Sophia also includes a virtual keyboard.

Sophia supports English, with additional languages to be added as requested by clients. Taptera is in discussion with a few companies, such as insurance and business consulting firms, according to O'Connor, about implementing Sophia for their sales teams.

Looking ahead, Taptera is also exploring the possibility of adding task management and reminder features to Sophia. "We want to respond to our customers' needs," O'Connor says. "If they're interested in developing a certain feature, we'll customize [Sophia] to their needs."

Not Quite Perfect

To find out how well Siri responds to various accents, TechRadar, an online publication, had staff members from Wales, South Africa, Germany, and Scotland ply Siri with several questions and posted the experiment on YouTube. In response to Donny from Scotland's question, "Who is the British prime minister?" Siri came back with "Sorry, I don't understand 'is the breath primester.'" Altogether, Siri answered 21 out of 36 questions correctly.

Although it has come a long way, voice recognition technology is not perfect. Since Siri was launched, for example, a number of online videos and articles have detailed the often-humorous mistakes the virtual assistant can make in response to verbal queries. So, naturally, Nuance, Angel, and Taptera will have to overcome this negative perception with their own technology.

Speech recognition is a "really tough nut to crack," remarks Forrester Research senior analyst Kate Leggett. "It's gotten incrementally better, but it's still not there yet. It takes a tremendous amount of effort in terms of maintenance and matching the speech patterns to get speech engines to work well."

Accents and variations in speech, combined with background noises, increase the chance that a virtual assistant will stumble when responding to users. "Dealing with accents and other factors in voice recognition is something everyone [in the virtual assistant industry] faces, and we continue to work on it," O'Connor says.

As for how useful these enterprise apps will be, "the jury is still out," according to Peter Voss, founder and CEO of SmartAction, a developer of artificial technologies for call centers. "If you're driving, for instance, you could ask [the virtual assistant] to give you the latest transactions over the past five days, but you would still be limited in terms of what you can do with that information," Voss says. "This is definitely an exciting technology, but how often…people [will] need to use a voice-enabled app is still a question."

Offering an incentive for customers to use new software can spur adoption rates, observes Shep Hyken, customer service consultant and author of The Amazement Revolution: Seven Customer Service Strategies to Create an Amazing Customer Experience."When the airlines introduced their online reservation systems, they offered incentives like extra miles or better rates," Hyken says. "Companies should give customers a reason to try something new, like voice-activated assistants."

For some customers, the idea that other people will see them using a voice-enabled virtual assistant is reason enough to try it out, suggests Lee Humphreys, assistant professor in the department of communication at Cornell University.

"There is a performative aspect here: What does it mean for me to be in public and ask Nina to book a flight for me, for example? Other people will hear me do that and it could be a sign that I'm busy and important," Humphreys says.

"This was very much the case [with] early mobile phone use," Humphreys continues. "It was adopted early by the business sector, and it was as much about conducting business as it was about performing."

There is also the question of privacy, especially when users may not want their mobile phone to blurt out personal financial data. "A key issue for us to wrestle with as a society is privacy, i.e., how much context these systems collect about us in order to better understand us and our needs," Lease says. "We might be amazed at how well the system understands us, or terrified by how little privacy we actually have today."

Either way, Lease admits, "voice-activated virtual assistants are already here and likely here to stay."

What's in a Persona?

When discussing Siri's new cohorts, it is impossible not to sound as though you are talking about a group of women: "Hey, have you seen Lexee, Nina, or Sophia yet?" This is not a coincidence, offers Lee Humphreys, assistant professor in the department of communication at Cornell University. "There are some interesting gender dynamics at work here," Humphreys says. "Personal assistants have been traditionally female, which seems to have been carried over [to virtual assistants." In the same way that most telephone operators were female, our history of receiving assistance from a woman's disembodied voice may have predisposed people to the idea of receiving help from a digitized female voice.

Another reason might lie in biology. "It's much easier to find a female voice that everyone likes than a male voice that everyone likes," Stanford University professor Clifford Nass, author of The Man Who Lied to His Laptop: What Machines Teach Us About Human Relationships, told CNN. "It's a well-established phenomenon that the human brain is developed to like female voices."

As for a virtual assistant's robot-like tone—despite her pleasant voice, no one would ever mistake Siri for a person—this could also be part of the app's persona, suggests Matt Lease, assistant professor in the School of Information at the University of Texas at Austin. "As a virtual assistant's behavior more closely resembles that of a person, it is more likely to disappoint us with its shortcomings," Lease notes. Developers can avoid this, he adds, by making sure the software system "makes no such pretense at seeming human."


Judith Aquino is an associate editor at CRM magazine. She can be reached at jaquino@infotoday.com.


Page1 of 1