Automated speech VUIs(voice user interfaces) have exposed us to a variety of greetings: Hey there! How may I help you? In your own words, describe the purpose of your call. Please press or say your account number. And the ever popular, This call may be monitored for training purposes. Well, at least most firms are getting better than the Your call is important to us, but not much.
November marks the beginning of the holiday season and a seasonal onrush of calls. It is the time when customer care, in general, and voice self-service, in particular, are put through the acid test. The annual spike in inbound activity brings proof once again that the phone is the primary instrument for electronic commerce and customer care.
If 2006 follows 2005's patterns, most shoppers will use Web sites and search engines to comparison shop before picking up the phone to find local merchants via directory assistance or to reach contact center agents to complete their transactions. After a sale, the telephone becomes the near-exclusive source of technical support and customer handholding.
Lest the automated speech community gets a swelled head over phone-based commerce outstripping Web-based activity, bear in mind that most of these calls are people seeking support, hand-holding, or other assistance from other people.
Does that make automated speech a matter of secondary importance? Actually, it's quite the contrary. Automated speech resources are the first thing that most callers encounter when they use the phone to reach customer care professionals. They are a means to an end, and the more quickly the customer care community recognizes where speech fits into their 360- degree view of the customers, the better they will be at providing callers with a gratifying and successful user experience.
It's About Success Rates
Whether it's conversational access to problem resolution, open-ended call steering, or the more arcane automated directory assistance, automated speech can be incorporated into almost every customer care call. In no way does this imply that 100 percent of those calls should be fully automated. The customer care community seeks the right balance of voice-self service and voice-assisted service.
The capture rate varies widely across vertical markets and specific applications. Something approaching 100 percent of balance inquiries for credit card issuers may be captured and kept in the IVR system or voice self-service platform. A growing percentage of those calls culminate in the automated handling of a "promise to pay" or "payment instructions" provided by spoken consent into a speech-enabled IVR system. Automated resources are positioned to fulfill a caller's original request quickly and then transform the call into an upsell or cross-selling opportunity. Such marketing strategies can work, but in the process might continue to build a level of resentment against automated speech systems in general.
Callers are demonstrably more tolerant of the boldface sales efforts from live agents than they are from automated systems.
There are direct parallels to the live versus automated battle going on among the new crop of directory assistance (DA) service providers that are offering free DA to callers in North America. As the demise of InFreeDA (provider of 1-800- 411METRO) indicates, the live versus automated debate has become a life-ordeath decision. Operator-assisted yellow pages and concierge services have provided some of the most visible examples of abject failures. MetroOne's InFone comes to mind, along with InFreeDA, but there are many others.
The free DA community is finding that advertising revenue and call-connection fees are too meager to subsidize the triple crown of operating expenses associated with promoting call volume, paying live operators, and recruiting advertisers. Automated speech technologies are crucial to keeping operating costs down. Nonetheless, it will be more important to encourage repeated use by responding to callers' queries in the most efficient and pleasing manner possible.
As the holiday season moves into full telephonic swing, let's all recognize that customer satisfaction and the perceived quality of any service call is not dependent on the accuracy of a recognition engine and robustness of the underlying grammar. Instead it relies on the quality of the underlying application, the accuracy of the data employed, and the breadth of information and services that can be associated with core query.
Dan Miller is a senior analyst for Opus Research. He founded Opus Research, Inc. and published Telemedia News & Views, a monthly newsletter featuring developments in voice processing and intelligent network services. Contact him at email@example.com.