On the Verge of Voice Commerce

Even though consumers are comfortable with voice technologies for things like directory assistance, demands for voice commerce in insurance have been modest. But that is likely to change.

Another emerging technology is voice over IP, or using the Internet to send voice data. Although it is not used widely among insurance companies, voice over IP, according to IBM's Sachar, could be a valuable tool for insurers, especially in the claims area. ""I am not aware of any insurance company using voice over IP with customers,"" says Sachar. ""There is no reason that it could not be part of a Web site. It enables people to collaborate over the Web-talk and share the computer screen when you have only one phone line. Combine voice over IP with other forms of technology such as a collaborative workflow application, and a real value begins to show. Applications can be shared-so two people can make changes to a document simultaneously. Applications like Netscape's (Mountain View, CA) Cool Talk and Microsoft's (Redmond, WA) Net Meeting are bundled with browsers and are in use today. I think that these technologies are of great value for insurance companies in the claims area, where collaboration is needed to resolve issues.""

Despite the advantages of combining voice over IP and other applications, customers may not yet be receptive to such a technology. ""The quality of speech with voice over IP is not as good,"" explains Sachar. ""The ear is very sensitive to the minor delays and irregularities of Web speech. Customers may find this a little bothersome."" Reflecting such concerns, Minnesota Life's Rossow says about the technology, ""we'll first focus on fundamental solutions like live chat, where information is typed back and forth.""

Although voice recognition seems to be the most popular of voice technologies among insurers, it is not for every company. Rossow researched the potential for use of voice recognition for Minnesota Life's customer service. Although he believes the technology is maturing, he fails to see it as a strong business driver for the term life insurance that Minnesota Life offers. ""I think that the nature of a transaction contributes to how useful voice technologies are,"" says Rossow. ""If you think about using voice recognition technology to service term life customers, there aren't many times that you need to interact with them. You may only have two to three transactions with those customers and then they die.""

Manulife Financial ($123.5 billion funds under management, Toronto), on the other hand, saw a voice recognition system as a necessary addition for its business. As the 401(k) writer added choices to its IVR system menu in order to accommodate more than 95 funds, the number of dropouts-customers who opted to speak to a live operator during the call-doubled. ""We knew we were either going to have to hire more service reps so people could speak to a live operator, or change the way that customers obtained their information,"" says chief strategy officer Kay.

Manulife chose to speech-enable the high transaction areas of its service offerings, like fund rebalance and inter-account transfers. The implementation process of InterVoiceBrite's (Dallas) IVR software bundled with SpeechWorks' (Boston) voice recognition software took between six and nine months, according to Kay, who plans to have the entire system speech en-abled later this year. ""We've re-ceived positive feedback from our customers,"" says Kay. ""We feel that this system is a business driver because it deals with a customer satisfaction level. People want access to know what their money is doing.""

Reliability of voice recognition systems varies across applications. ""At Nuance we measure performance on accuracy,"" says Nuance's Ehrlich, who explains that reliability is relative to the size of the list of recognizable words in a system. ""If a system must recognize 20,000 stocks and 3.5 billion ways that they can be said, that is much harder than having a system recognize 'yes' or 'no'. I'd say on average there is about a 95 percent accuracy rate and that number may be slightly better or slightly worse depending on the application."" According to Ehrlich-the number which is typically in the 90 percent range-is as good as or better than what a human can recognize. ""If I were to tell a person that my account number is KSDLH, often a human will have you repeat it with questions like 'Did you say B or D?"" adds Ehrlich.

2 of 4