How Speech Recognition can Improve Customer Service

By Paul Kowal

Convenience and personalization are the hallmarks of good customer service, but customers don’t often experience both on any given call.

Click here to enlarge image

The two predominant methods of customer service–touch-tone multiple-choice menus and phone representatives–are limited in their capacity to provide superior service. While they are convenient, touch-tone multiple-choice menus are an annoyance to many and usually limited in their ability to provide answers to anything beyond the most common questions. On the other hand, phone representatives provide the personalization and depth of service that customers really appreciate, but calling during business hours Monday through Friday isn’t usually convenient.

With the two predominant methods coming up short in offering full customer service, it is no wonder that businesses–utility companies included–are looking for alternatives.

Speech recognition, the ability of computers to understand and process spoken language, may soon become a popular alternative for utility companies. It is a smart and effective phone center service solution that is available at a fraction of the cost of a team of customer service representatives. The computerized system “listens” and responds, 24 hours a day, seven days a week. This “conversation” is carried out in a friendly manner with speed and efficiency.

Over the past several years, speech technology has evolved, and it continues to improve. Early applications of speech recognition systems were often clumsy and impersonal. Callers had to over-annunciate to avoid being misunderstood. Many of the speech recognition systems had trouble with accents, foreign languages or speech patterns, and struggled with pitch. This posed a problem for speech recognition developers because, if the people using the system could not be understood, they might become frustrated or insulted and not call back. Due to the dramatic increase in both accuracy and vocabulary, this has all changed.

Recently, speech recognition reached another level in its evolution: personalization. Through some simple programming and a good text-to-speech engine, callers will perceive that the application has been personalized for them.

For a utility contact center interested in deploying a personalized speech recognition application to improve customer service, the benefits are many–and the costs are small, compared to the benefits. Calls are shorter so more calls can be handled, improving customer service. More importantly, a recent study found that 74 percent of callers who have used a speech technology application found it to be as good as a live customer service representative.

“Personalization, if done right, enhances a caller’s experience,” said Blade Kotelly, worldwide marketing solutions manager for SpeechWorks, a provider of speech technologies and services. “And a good customer experience creates great branding.”

Here’s how personalized speech recognition works: The system is programmed to respond to the existing data, which is based on the habits, routines and the demographics of callers. With each call, the computer gathers and stores more information to be used for customizing future responses. The data is read through a text-to-speech engine.

For example, a personalized speech recognition system programmed to respond to habits and routines of a caller could sound like this:

*System: “Thank you for calling again. Would you like to check your current balance today?”

*Caller: “Yes.”

*System: “Your balance is $12.03.”

Without the personalization feature, the call would have been longer and would require that the caller answer multiple questions before receiving the information they wanted.

Kotelly, author of the upcoming book, The Art and Business of Speech Recognition: Creating the Noble Voice, said, “First-time callers may answer more questions as the system collects the necessary information. They’ll be asked the opening question: ‘What would you like to know?’ and take it from there. But that question won’t pop up again because, after each call, the interaction becomes more personalized.”

There are a variety of techniques and features that can be implemented to customize each caller’s interaction. One is the use of skip lists, which enable the application to skip options that are not appropriate for a particular caller. Once a question has been asked and answered, the information is stored so that the next time, that prompt or question is skipped. This makes for a more efficient and personalized calling experience.

“Skip lists are an important feature,” Kotelly said. “There is such a thing as too many questions. Customer-service-driven speech technology applications are dedicated to avoiding the tedium of touch-tone multiple-choice menus. Smart programming keeps speech recognition from becoming monotonous. You don’t want to waste the caller’s time. More time equals less value.”

Speech recognition systems are offered in different languages, and can be customized to identify different accents or dialects of a specific region. Again, people will feel the system is personalized to them and their needs. Callers can simply say, “I would like to speak in Spanish during this call.” Not only will the system switch to accommodate this request, but it will also store the information so that the next call with this customer will be in Spanish.

Another technique used to personalize speech technology is matching confirmations. This means that the application adapts its speech pattern to mirror the caller. For example, if a caller says “o” instead “zero,” the system will match the manner of speaking and say “o” as well. If the caller says “IBM” then the application will respond in kind, saying “IBM” as opposed to “International Business Machines.” This feature adds to the natural and more human flow of the conversation.

Randomized prompts are another feature that round out the human feel to speech recognition. These prompts are programmed so they communicate the same general message, but are slightly different with each question. For example, the computer might say “Thank you for entering your account number. Just for confirmation, is it 4293587931?” The next time the question might be rephrased, and the system could state, “4293587931 is the number I heard. Is that your account number?” By changing the prompts, the interaction is more conversational and more personal. Talking to a machine reading a script is equally as unsatisfying as speaking with a person reading a script.

Overall, these unique features and personalization techniques add up to a more enjoyable experience for the caller. As a result, utility companies that use personalized speech technology will stand out as a warm and friendly. Personalization of speech recognition systems may not only improve the bottom line for companies, but it might save consumers from touch-tone torture and antiquated robotic systems.

G. Paul Kowal is founder and president of Kowal Associates Inc., a contact center consulting firm that has been helping Fortune 500 clients maximize the effectiveness of their multi-channel customer contact strategies since 1988. Paul has been the lead consultant to Fortune 100 companies regarding teleservices and customer service for a variety of CRM programs.

Previous articlePOWERGRID_INTERNATIONAL Volume 8 Issue 3
Next articleEL&P June issue features Industry Report on merchant development

No posts to display