Anybody's Bot
The definition of “virtual agent” seems to vary depending on who you’re speaking to. Even the top three virtual agent companies have different opinions when it comes to what capabilities ultimately define a virtual agent.
Mark Gaydos, vice president of worldwide marketing at VirtuOz, says that VirtuOz defines a virtual agent as having “the ability to interact with a customer and, in a sense, understand what the customer wants to do and guide them to an answer to their question.” According to Gaydos, the understanding component is what keeps Microsoft’s signature paperclip help feature, Clippy, from being considered a virtual agent. “[Microsoft was] showing an avatar on top of a search engine and calling it a virtual agent,” Gaydos says. “[Avatars over search engines] have enough semantics to understand and know that ‘restaurant’ means ‘diner’ means ‘coffee shop’ and it can bring results that are wider, but it really isn’t having an interaction or a conversation [with the customer].”
David Lloyd, CEO of IntelliResponse, agrees, explaining that a virtual agent “is a way of interacting and delivering customer self-service.” However, IntelliResponse stresses the search component as a major ingredient in providing customers with what they need, especially in an age in which many businesses have inadvertently overcomplicated their Web sites with too much information.
“[IntelliResponse’s] virtual agent platform allows our customers to rapidly implement a virtual agent that can deliver one right answer in mobile, social media, agent chat, agent voice, agent email effortlessly,” says Mike Hennessy, vice president of sales and marketing at IntelliResponse.
eGain gives considerable more weight to a virtual agent understanding text and speech input by a customer, as explained by Ashutosh Roy, chairman and CEO of eGain. “Our definition would be [that] a virtual agent is a human-like bot that understands natural language interactions and guides a visitor though a conversational interface using text and/or speech input and output,” Roy says.
To further complicate terms is the differentiation between virtual agent and “virtual assistant,” as eGain referred to its virtual agents as virtual assistants before renaming them as “chatbots” in 2007.
Diane Clarkson, online customer service analyst at Forrester Research, maintains that these variations are simply a matter of semantics, as virtual agents only differ slightly.
“Many [virtual agents] do similar things but [companies] call them something different because people were dissatisfied with virtual agent. Some of them are called something different because [businesses] didn’t like the term,” Clarkson says. “I think the word agent doesn’t always work for people. ‘Virtual’ has an X-box connotation. ‘Agent’ doesn’t always suit the terminology for the industry.”
Clarkson points out that while some customers find the word agent intuitive with regards to some industries, such as travel agencies for example, others feel more inclined to alter the term to attract customers.
Virtual Results
Regardless of terminology, virtual agents are, across the board, being received well by customers of many industries. One of eGain’s U.S. customers, for example, concluded that it was able to offload 30 percent of customer queries at a 96 percent resolution rate.
In the Bloomberg Businessweek piece, “Virtual Agents Will Replace Live Customer Service Reps,” Gaydos wrote, “Virtual agents are already augmenting and replacing human support personnel. You don’t need fancy arguments to explain this phenomenon (we have built computers to beat chess champions, so having them help customers with bill disputes is a walk in the park).”
VirtuOz determined that SFR, a division of the mobile communications enterprise Vodafone, uses a virtual agent to facilitate 750,000 conversations a month regarding customer account questions and the company’s services. eBay, one of VirtuOz’s customers, is already facilitating 200,000 customer conversations a day with virtual agents in six different countries.
Virtual agents are effective in dealing with specific customer questions because of what Roy refers to as the 80/20 rule. This standard distribution relies on the fact that 80 percent of the questions asked by customers revolve around only 20 percent of the topics. Although this logic has been applied to generating FAQ lists, Roy asserts that this reasoning can be applied to virtual agents as well.
Because so many businesses now strive to make personalized experiences for their customers, virtual agents can also provide a more cost-effective way to tailor experiences for consumers. “[Businesses are] creating experiences that are more personalized, less mechanical.... [There is this] same notion of, ‘How do you use technology to create a perception of intimacy—of familiarity without spending a lot of money?’” Roy poses. “A chatbot is one of the tools in that exercise. It’s a tool that creates a perception of human-like interaction, if blended well. [Virtual agents] need to blend well with human conversation. Over time, what you want is that the visitor should not be able to tell the difference between the virtual assist or a human being responding in a seamless conversation.”
Growing Independence
Analysts and vendors agree, though, that what most appeals to customers about virtual agents is the opportunity to be more self-reliant. In a recent Harvard Business Review blog, “Why Your Customers Don’t Want to Talk to You,” Matt Dixon and Lara Ponomareff determined that three to five years ago, two-thirds of customers primarily used the phone for service interactions. At present, less than a third do. However, of that one-third, 57 percent first attempted to resolve their issue on the company’s Web site. And, in addition to that, over 30 percent of all callers are on the company’s Web site at the same time that they are talking to a representative on the phone.
“I think people like to get answers to their questions and they’d prefer not to interact with a person. Most people want to be self-reliant. There is an obligation that goes along with interacting with a person. Most people want to do their own thing,” Gaydos observes.
Younger customers, in particular, are demonstrating a trend of perusing the Web for their own answers to inquiries, opting to search Google or forums rather than pick up the phone. With this trend comes a tremendous opportunity for virtual agents to assist in the process of self-service. “As the younger generation becomes more influential in terms of overall spend, companies will have very robust self-service, particularly on mobile devices and possibly on in-store kiosks,” Clarkson says. “As virtual agents become more widely adopted by Web surfers, the skew towards younger will shift.”
Lloyd agrees, pointing out that this trend makes sense considering the younger demographic isn’t tethered to a laptop or an office. Most college students are choosing to resolve problems via their mobile devices, remaining stationary only for the moments it takes them to check Facebook, which, incidentally, can also be done remotely.
Athabasca University, a Canadian institution specializing in distance education programs and courses, has been well aware of its students’ mobility and, in 2003, began using IntelliResponse agents to address the growing volume of requests in its Information Center. In what it has aptly named AskAU, Athabasca University has already successfully integrated virtual agents into Facebook and iTunes using IntelliResponse applications. But in February 2010, Athabasca University’s internal development team went a step further and created Sunny Davros, a virtual Second Life agent who inhabits AU Island.
“The creation of Sunny Davros in Second Life is an extension of work with conversational agents that had been conducted at AU over the last five years or so,” says Michael Shouldice, senior recruitment officer of Athabasca University. “As social media began to grow we, as did IntelliResponse, recognized that it was important to be a part of the conversation. When IntelliResponse offered the Facebook and iTunes app functionality it only made sense to include these offerings. Our students, who study with us around the world and around the clock, now have a number of access points to seek answers to their questions about AU when and where they want to.”
Behavioral trends aside, the “noisiness” of so many company Web sites is causing customer self-reliance to become more prevalent in the older demographic as well. Lloyd concurs, admitting that “The younger [demographic] has a shorter attention span, but it’s not a situation that is [necessarily] relegated to the younger demographic...this is only going to spread more rapidly.” Lloyd points to the growing demand for smartphones, a product that is sought after by consumers of all ages.
Because many company Web sites have become so overwrought with information, Roy argues that “virtual assistants have the potential of being the concierge through the online experience for all demographics.” According to Gaydos, many companies have unintentionally made their Web sites more complicated than ever, adding endless pages of information to accommodate questions. This tactic has, in turn, made Web site navigation much more difficult for customers hoping to locate a specific answer.
“On your Web site, no one can hear you scream,” Lloyd adds humorously when describing the frustrating experience of mining through Web sites.
Gaydos also uses the word concierge when explaining the role of the ideal virtual agent, adding that customers should not be inundated with so many details, loopholes, and special circumstances regarding their inquiry. “The robots are here to help you and they have warm hands,” Gaydos says. “You don’t want the customer to know all that information.... [The virtual agent should] figure out all the rules so the customer doesn’t have to. That’s what [VirtuOz] means by ‘shepherding the customer.’ [VirtuOz] hides the complexity from the customer.”
Gaydos uses the metaphor of choosing the perfect wine to illustrate his point: “It’s the difference between giving someone a wine list and asking them to choose and someone who talks to you and asks you, ‘What’s your price point?’ It’s about honing in on those key things that is really problematic for your customers.... The information is on the Web site. They just can’t find it.”
To Bot or Not?
Although virtual agents should be able to guide a customer through more involved questions, the point at which a human agent should be reached is uniformly acknowledged as the big three: the cancellation of service, cross-sell or upsell opportunities, or to deal with a high-value customer.
However, Clarkson notes that there are some small exceptions when it comes to cross-selling, specifically regarding higher-volume, lower-volume revenue. She uses the example of an iPod and an iPod skin when she says, “Higher-volume-lower-revenue-additional items can be very important to improving a customer’s experience.” A customer who is buying an iPod online might appreciate being offered a multicolored skin to accompany that purchase, an offer than can be entrusted to a virtual agent primarily because of the low cost of that item. “[That] customer will walk away with a good experience,” Clarkson adds.
Although many customers are becoming increasingly more receptive to this type of self-service, phone calls still have the highest resolution rate over any other channel. This schism between customer behavior and problem resolution can be explained in a customer’s willingness to go with the faster option.
“The big issue with the phone is that it isn’t necessarily faster than self-service or chat, particularly for low or medium complexity issues,” Clarkson says. “It requires getting to the right person, being on hold, [and] having to explain the situation. The phone also isn’t as private as self-service, particularly for consumers who are multitasking and trying to resolve customer service issues at work.”
Analysts and vendors agree that a customer engaging with a virtual agent should have ready access to a human agent, should the customer’s satisfaction begin to wane.
“[eGain’s] approach here is that we have a common platform that provides a core set of services around a common knowledge base,” Roy says. “The way it works is the virtual assistant is set up to escalate depending on configuration. [You can] either go to a chat channel or a click-to-call phone conversation, and you can set it up so that there are [specific questions, such as], ‘What’s the price of your software?’ That’s a typical escalation product. Escalation can be far out of the domain of the chatbot. It needs a human assistant.”
One company that uses eGain software always keeps a human agent nearby, often keeping a close eye on the exchange between virtual agents and users.
“If the chatbot is not able to resolve a query to a customer’s satisfaction,” explains the customer, “it offers to escalate. eGain’s unified multichannel Customer Interaction Hub allows seamless, context-aware escalation to an agent (e.g., chat agent), where the chat agent is fully aware of the customer interactions with the chatbot and does not need to repeat the same questions.”
This constant monitoring of virtual agent and customer exchanges is the most efficient way to use virtual agents, Roy says. He advocates an eventual indistinguishable voice between virtual agent and human agent that should go undetected by the customer, punctuated with proactive offers, of course.
Future Bot Generations
Clarkson deems virtual agent technology as “very forward thinking,” but points out that the future of the industry will go hand-in-hand with language processing. “A virtual agent is only going to be as good as the natural language processor, which is related to the vendors,” Clarkson states.
Roy agrees, explaining that to his knowledge, all vendors currently lack the speech interface to create a truly seamless experience. “You want to be able to just chat to this bot and the bot should be able to respond with speech and take you to the right place on the Web site,” Roy says. “That piece is still missing. That interface needs to be more natural.”
Another component that needs to be constantly improved upon is the knowledge of the virtual agent. In assisting customers with an ever-expanding collection of questions, virtual agents must be prepared for any new developments, no matter how minor. “You have to maintain the knowledge of the virtual assistant,” Roy advises. “There’s no easy way for the chatbot to automatically leverage existing knowledge in the system. It’s very frustrating. It starts to bog you down because you have to keep updating the know-how of business systems and that’s a gap that I see in the second generation. Like eGain, providers are starting to address [this need].”
Another challenge for the virtual agent industry is keeping the agent constant across multiple channels, a challenge that Gaydos says is necessary for the evolution of the industry. “The way virtual agents are going to go is that they’re going to get more intelligent, more pervasive,” Gaydos envisions. “You’re going to see them appearing consistently across channels. [Right now], there is a huge channel conflict problem...you want that consistent concierge. No one has nailed this that I know of.”
Similar to Roy, Gaydos expects to see virtual agents with “human overlords,” overseeing these many conversations and intercepting when necessary; Gaydos also expects to see virtual agents communicating with other virtual agents, sharing customer information and the context of inquiries.
Clarkson believes that search site boxes will eventually become extinct along with FAQs due to the virtual agent. “I would see virtual agents becoming much more crucial to the online self-search function. [Virtual agents would become] much more sophisticated and replace a lot of self-service content...it could become the core part of the self-service help section,” Clarkson explains.
Others speculate that virtual agents will complement conventional customer interaction as they become more popular, particularly with voice interaction. “There’s going to be an inflection point largely around the speech-based experience,” Roy predicts. “You visit a Web site on your smartphone and if you have a question you can just speak your question...[like a] personal assistance application.”
Yet, when Gaydos is asked if virtual agents will eventually come to replace human agents altogether, he says, “they already are.”
Editorial Assistant Koa Beck can be reached at kbeck@destinationCRM.com.