Early adopters report benefits with new Microsoft Speech Server

They see significant benefits from the technology in the years ahead

SAN FRANCISCO -- Seeking to push voice recognition technology into the mainstream, Microsoft Corp. last week pointed to encouraging signs as it launched its Speech Server 2004 software here.

Early adopters that have worked with consultants using beta versions of Microsoft's maiden product either have gone live with applications that can recognize speech responses or will soon move beyond the pilot phase. They said their experiences so far indicate that they will see significant benefits down the road.

Seattle-based Grange Insurance Group predicted a 15% annual return through a reduction in the number of calls its customer service representatives handle once its 150,000 policyholders get the chance to use the new system. The company is currently piloting a speech-enabled application that lets 750 policyholders check billing information over the telephone by recognizing their responses to questions. Grange is also letting mortgage companies access additional policy information.

CIO Ralph Carlile said Grange had planned to make the information accessible through Web-based applications. He said that once he learned that his company could reuse the development work to make the information accessible through the telephone by speech-enabling the applications, he jumped at the chance to participate in Microsoft's joint development program for Speech Server.

"A lot of policyholders don't have Internet access or fast Internet access. A lot are older people who are not comfortable doing anything on the Web, but they are very comfortable on the telephone. This is almost a no-brainer for us in terms of a third channel for doing business with our policyholders," he said. The company plans to extend the system to enable payments to be made via the Web or telephone.

Grange worked with Tata Consulting Services Inc. on its initial application, although Carlile said his company's in-house developers will be able to do 50% to 60% of the work on the new project, which he estimates will take two months. "Once the infrastructure's there, doing development for subsequent applications and solutions is actually quite rapid," he said.

The rationale for adopting Speech Server was similar for the Southwest Alabama Integrated Criminal Justice System. SAICS used Microsoft's Visual Studio .Net development tools to build applications that provide Web browser-based access to driver's license, Social Security, license plate and other information. Since police officers on bicycles and patrol boats don't always have access to computers, the prospect of getting information through voice inquiries on cell phones was appealing to Jim Pritchett, executive director of Foley, Ala.-based SAICS.

"Since our original system was all based on Microsoft .Net technology, our query engine was already there. We already had the Web services built. We were able to take the development tool, Visual Studio, that we were using and initiate the speech application just by building the linkages," he said. "We don't have to train anybody on anything different. We use the same queries that we had already built. We just are adding a new layer of functionality to that existing query."

SAICS worked with Toronto-based ComputerTalk Technology Inc. on the application, which went live at the end of February. Pritchett said SAICS is already seeing a reduction in the volume of radio calls to dispatch centers.

The customer examples Microsoft promoted caught the attention of several attendees at last week's developer conference.

"It's very cool. I just don't know yet where we'd use it," said Mike Jacobson, a programmer/analyst at Labor Ready Inc. in Tacoma, Wash.

"I don't think it's the end of human interaction, but it gave us lots of ideas. I think eventually we'll get to it," said Javed Fouch, a programmer at Hagerty Insurance in Traverse City, Mich. By automating calls for basic information, his company might be able to curb the growth of its customer service department, he said.

Microsoft is promoting the price point of its speech-recognition technology to set its offering apart from competitors such as Nuance Communications Inc., ScanSoft Inc. and IBM. But one prominent user pointed out that price isn't the most important criterion when weighing speech-recognition technologies.

"The most important thing for me in any speech application is the power and the strength of the speech recognition," said Fari Ebrahimi, an Arlington, Va.-based senior vice president in IT at Verizon Communications. "What matters is, Are we going to give customers an experience they'll really enjoy and [make them] feel that we cared about them? There's no price to put on that."

Ebrahimi said he will look at Microsoft's Speech Server, since Verizon uses other Microsoft technology and he likes the notion that various components are integrated with the Microsoft product. He said the company might opt to stick with its ScanSoft speech-recognition engine, which can plug into Microsoft's Speech Server.

Cost was a factor for SAICS, which piloted Nuance's product before opting for Microsoft. By using its familiar Visual Studio tools, Pritchett said, SAICS was able to develop and deploy its application using Speech Server for at least half the cost and in a quarter of the time. The biggest cost associated with the Nuance-based system was the licenses, he said, noting that price comparisons can be tricky, since SAICS got favorable treatment with Microsoft's early adopter program.

Carlile said that he has been impressed by the performance of the Microsoft-based product and that Grange has seen "very little failure rate." But he added that he hasn't done a side-by-side comparison against any other speech-recognition products.

Art Schoeller, an analyst at The Yankee Group in Boston, said Microsoft's speech-recognition technology will need more time in the field to rise to the level of the market leaders. But its price could make it attractive to small businesses.

"Large businesses who focus only on the drop in speech-engine license fees are missing the point about deploying good speech apps," he said. "You still need folks experienced with good voice user-interface design. We made the mistake in the mid-90's, telling folks they could easily develop touch-tone-based applications. Some were good and some were very bad."

It's not clear how soon Speech Server will have an impact in the market, analysts said.

"They've got a good bunch of early adopter customers and [joint development partners], but it's unproven in terms of large-scale and wide-scale deployments," said Steve Cramoysan, an analyst at Stamford, Conn.-based Gartner Inc. "They have a limited number of customers, so they can focus their attention on making sure those deployments go well."

Dan Miller, an analyst at Zelos Group in San Francisco, predicted that Microsoft's market share will slowly build as the marketplace puts more emphasis on enterprise middleware rather than on traditional voice ports. "We're going to start looking at the world in a different way," he said.

The speech technology market is expected to expand, regardless of Microsoft's entry into the space. Datamonitor PLC estimates that revenue associated with the voice business will grow from $806 million in 2003 to $1.3 billion by the end of 2005, according to Daniel Huong, an analyst at the company. Huong said the improving global economy will free up IT and marketing budgets to permit further investment in customer care and self-service automation.


Copyright © 2004 IDG Communications, Inc.

Shop Tech Products at Amazon