What's new in speech recognition for the call center these days? Nothing, says Aspect Software's director of business process marketing Thomas Chamberlain, and that's new. "I think the trend that we're seeing is that it's no longer a new technology - it's really hit the mainstream. The technology has evolved to such a point that it's no longer early adopters," he told us.
We're in a stage where call centers are discovering the relative advantages and disadvantages of hosting versus premises-based systems and directed dialogue versus natural language. Some, like Angel.com's CEO Mike Zirngibl, stake their whole business model on directed dialogue and hosting, but most see a more organic blend of hosting when it's cheaper, premises-based when security calls for it and natural language when the options are too complex and open-ended for directed dialogue.
The Hosting Question
Last April, we asked some speech recognition service providers if they thought hosting was the future of speech. As hosters, their answer was, naturally, a resounding "yes!" Were they right? "They have a point," says Datamonitor analyst Daniel Hong. "Right now hosted solutions are more attractive because of the lower entry costs for speech but with the growing sphere of developers around Voice-XML and speech in the next five years premises-based solutions will also have a strong future. As a vendor it is essential to have flexibility across different deployment models."
But the hosting trend is exciting because it means that some very sophisticated technology is becoming more available to smaller centers.
Brian Garr, IBM's program director for enterprise speech solutions, told us: "We believe that hosting is the fastest growing segment of the speech ecosystem because it reduces the capital outlay required for medium sized companies to get the hardware and software in place to be up and running with speech self-service applications."
Even Aspect Software, a company that serves larger clients that have the infrastructure and personnel to handle huge premises-based systems, admits there's a place for hosting: in handling seasonal spikes in call volume, to be exact. "Primarily we see it in areas where they have cyclical or seasonal variations in volume," says Aspect's Chamberlain. "[Call centers] don't necessarily want to purchase capacity in their premises solutions for a spike but they would basically use it as an overflow to a hosting solution just to get by the particular cyclical or seasonal volume change."
The primary reason not to go hosted? Security. Many of Aspect's customers, those in healthcare and financial services in particular, find that letting sensitive data pass through an outside firm is not worth the risk.
Peter Mahoney, Nuance's vice president of worldwide marketing for speech solutions, observes: "We believe that most enterprises want to maintain a choice in their deployment model for speech applications. The vast majority of large enterprises deploy speech applications on premises today. We expect that some portion of our customers who are deploying speech applications inside their own IT infrastructure will choose to move all or part of those applications to a hosted model."
Natural Language
IBM's Garr says of natural language speech recognition that "the advantage is the ability to navigate using natural utterances. One disadvantage is that the customer may simply not know what to say when not 'directed.'"
Skeptics may ask why we need speech recognition when touch-tone will do. Furthermore, if directed dialogue speech recognition works, why bother with natural language systems? Good questions. A common mistake in speech implementation is merely exchanging your touch-tone system for a speech-enabled one. Aspect's Chamberlain: "The reality is if you take a touch-tone and just basically put speech on top of it doing the exact same thing -- what a mistake. You've changed your customer experience with no value-add. If you're not going to add value for your customers - and they'll recognize that right away -- your self-service rates, which affect your cost structure, will go down."
And you have to know when to use speech. Natural language systems are perfect for routing, notes Nuance's Mahoney. "More simplistic applications -- like simple information retrieval tasks," he adds "can often be more economically developed without the use natural language techniques."
Robin Biddy, vice president of sales and corporate services for CPT International, a speech application hosting firm, says that natural language speech systems can cut down on call durations because they take less time for callers to get to the information they need. But, Biddy adds, "With natural language applications it is critical that the business rules that govern the application are clear and agreed upon before the application is designed. Even more than converting from a touch-tone to a directed dialog application, natural language requires truly innovative approaches to application design to ensure successful results for the business and their customers."
Caveats
Perhaps the biggest mistake businesses make with technology is using it because it's there. Don't replace your touch-tone system with speech just because you can. Aspect's Chamberlain said it in the previous section -- don't drop a speech-enabled system on top of your touch-tone system.
Nuance's Mahoney warns, "When defining a set of applications to automate using speech, some clients feel compelled to take their existing touch-tone applications and convert them directly to speech. Since speech enables brand-new kinds of interactions, it's critical to look at the application and define the best way to solve the customer problem without retrofitting the application to behave like the older, more limited technology."
CPT International's Biddy put it simply: "Some applications just don't lend themselves to speech."
If you do decide to go with a speech-enabled system, be sure to design the voice user interface carefully, says Brian Garr. "A bad VUI can actually cause a loss of customer retention and cause more calls to roll to an operator."
But don't assume callers will interact with the speech system the same way they would with an agent or a touch-tone system, says Convergys's director of business development Steve Chirokas. "Call centers considering a speech system should take the time with an experienced application development resource to properly scope the application, understand the impact of the persona, test the dialog and spend time to tune the system."
An Unholy Alliance? Microsoft, Nuance, and Paul English
Paul English, the blogger formerly known as the scourge of the IVR industry, is being sucked into the Microsoft machine, lending his crusading Gethuman.com website as a forum for a push for user-friendly IVR standards.
English made headlines last fall with his "IVR Cheat Sheet," a list of short-cuts for callers to bypass cumbersome phone trees to connect directly with call center agents. Mr. English decried what he saw as the industry default: keeping callers away from agents by forcing them to exhaust all automated possibilities first. English's solution was to show the public the "secret codes" -- like hitting the zero key over and over or repeating a word -- to get straight to an agent.
Although some call center industry insiders didn't buy it, Mr. English maintained that he wasn't against IVRs across the board -- just poorly designed systems. Now English has made good on that by hosting a standards forum on his website, www.gethuman.com.
The standards forum would have little weight without the power of Microsoft and Nuance behind it. But it's also a great way for Microsoft and Nuance to make public relations lemonade out a story that has been a bitter lemon for the IVR industry.
VocaLabs' Peter Leppik wasn't convinced by the software vendors' good will. He wrote on his company's blog: "Not to be too cynical here, but in the meanwhile, Nuance and Microsoft both get free PR about being "good guys" and promoting quality customer service, but when push comes to shove they'll still build crummy systems if that's what their clients demand."
Sam Aparicio, one of Angel.com's founders, reserved his skepticism for English. He wrote in his company's blog: "I find Paul English a much better problem finder than solution proposer. His message resonates because some large companies have stopped being customer-centric, and we are all customers of those companies."
But now English has some solutions to propose. What are we to make of them? In his summary of the situation, Opus Research senior analyst Dan Miller wrote: "Implicit in English's recommendations is the idea that reaching a live agent is the solution to caller's lack of empowerment and overarching frustration." Miller thinks that the English/Microsoft/Nuance standards list will "by definition, only address a fraction of the user experience problem." In Miller's estimation, most of the problems callers complain about have, at their source, staffing, training, and business culture to blame.
The first version of the standards list included typical suggestions about agents answering phones when available, a standard zero-out to agents, frequent wait-time estimations, and others. For the rest of the standards list, go to www.gethuman.com/earcon/standard.html.
We asked Thomas Chamberlain of Aspect Software what he thought of the standards push. "A 'zero out to operator universal' -- as a matter of fact, it's almost universal today. What's new about that?" What indeed?
The standards list was scheduled to be up for August and September to give the industry and the public time to comment. After that, "Companies can then register their GetHuman-compliant phone service at www.gethuman.com and begin to adopt the GetHuman 'earcon,' an auditory icon that signals to callers that the company uses best practice customer service standards," said a statement released by the three groups.
http://www.callcentermagazine.com
Copyright 2006 CMP Media LLC. All rights reserved.
10/1/06, Issue # 1910, page 22.
