The Answer to Customer Service?

Many times, customers experience a trade-off when they try to find satisfaction through machines: what they gain in convenience they lose with modern-day annoyances that come with the new technology. Today, you can check your savings account balance at any time of the night, but not until having to cope with about five rounds of touch-tone multiple-choice questions. You can order clothing online, but if there are delivery problems, e-mailing customer service can be frustrating, especially if you have to wait hours--or days--for a response. For years, speech recognition technology was viewed with a more critical eye. Not only was the technology seen as clumsy, with people having to overly enunciate and use a very limited vocabulary, but many times, the technology simply didn't work properly. The concept of talking to a computer or telephone and having it (accurately) respond to your commands was seen as something for science fiction. Passed up as unreliable a decade ago, speech recognition today receives high marks for its high completion rate, accuracy and large vocabulary. Companies are so confident of speech recognition's power that the technology now manages customer service interactions. This is no small task in a culture where consumers expect next day delivery and the ability to do business at any time of day or night. But according to the Gartner Group, "Speech recognition technology finally works. (It is) an emerging 'self-serve' technology that will enhance customer service while reducing personnel costs." Speech recognition software and hardware have the capability to change the way we think of customer service. While customer service today is dominated by Dual Tone Multi Frequency or DTMF systems that require the tedious use of touch-tone phone keypads to plod through predetermined branches, tomorrow belongs to the ease, efficiency and cost savings of speech recognition. Here are several reasons why speech recognition is so well suited for customer service: EASE OF USE
A real company using DTMF customer service actually provides the following outline to help customers who wish to change their PIN numbers using their touch-tone phones: Press 1, then 4, then wait for the dial-in welcome, then press 2 and star. Then press one and star. Enter your (personal code) and star, then press in your PIN and star. Then press 6 and star. Now you can dial the phone number and start the procedure to change your PIN number. Using this application, customers would have six rounds of button pushing before beginning to do what they want. Is it any wonder that 60 percent of customers using these systems have hung up on them before completing their tasks? While convenient, DTMF systems are notorious for their tedious routing. Sometimes, tasks that would take less than a minute with a live phone rep can be frustrating and time-consuming as customers are subjected to numerous "multiple choice" options. With a speech-enabled system, a customer could have duplicated the above procedure by simply stating, "I want to change my PIN number." Instead of choosing from a menu which may not list their needs, consumers just say what they want. Furthermore, unlike DTMF systems, speech technology can be used from both touch-tone and rotary phones. And it is much less awkward to use speech when calling on a cell phone, or even a cordless handset at home, since the keypad is in the same handset you need to hold to your ear to hear the prompts. IT'S WHAT CUSTOMERS WANT
When it comes to customer service, customers say they want efficiency, speed, accuracy, convenience, friendly interaction and a system that is easy to understand. Speech recognition technology can fulfill all of these needs - even "friendly interaction." Companies developing speech recognition systems seem to have adopted a much more casual, conversational speaking style in their prompts, showing that it is possible to achieve friendly interaction without a live call center representative. Companies prefer speech recognition software and hardware for the same reasons they rushed to install DTMF IVR systems 10 years ago: It is cost-effective, it enables their customers to use it at any time and it delivers a consistent message. But unlike DTMF, consumers also appreciate the new technology. In surveys, over 80 percent of consumers are at least somewhat satisfied with speech, and two-thirds claim that speech is as good as a live customer service representative or better. In one survey with a high tech audience, 60 percent of those surveyed said that they actually prefer being handled by speech technology to CSRs. IT'S COST-EFFECTIVE
When implemented, a speech recognition program can cost 90 percent less than a human CSR. No, that isn't a typo. Here are the figures: combining salary, recruitment and management costs, as well as the expenses of facilities and benefits, a CSR costs an estimated $34,800 a year. With a speech port in a medium-size system, the installation, maintenance and overhead costs total $3,333 a year. Broken down by call, customer service costs $1.47 per call through a CSR, and 9 cents for a speech recognition port, 16 times less expensive. All for a service that 60 percent of customers say they prefer. POSITIVE IDENTIFICATION
One of the biggest limits of DTMF technology is its weak security. Because of their structure, DTMF systems rely on PIN numbers that can be easily forgotten or stolen. Many times, a person only needs an account number and a few Social Security digits to have full access to someone's savings or brokerage account. Speech recognition can combine understanding with security by creating a voice print of the caller. A customer only have to "register" once using his or her voice, and the voice pattern is stored in a small file for any future transactions. When the customer calls the system again, he or she needs only to say a few key words (usually a natural element of the call) to gain full access to the account. Each voice pattern is unique, and can't be duplicated or changed. Even if the user has a sore throat, the customer will still be recognized. Speech recognition eliminates the need for tedious, extra steps in the identification process (e.g., "Can you tell me a recent transaction on your account?") that are unreliable to begin with. In fact, speech recognition systems can be used as a two-tiered security system, prompting users to speak their account numbers or dates of birth, and simultaneously checking the validity of both the information and the voice. If there are any discrepancies, the caller is routed to a live representative. A PROVEN TECHNOLOGY
Speech recognition is now being used with phenomenal results. Today, Fortune 1000 companies such as Sears, Hewlett-Packard and Federal Express use the technology to give customers technical help with computer problems, trade stocks and provide 401 account information. In the case of Federal Express, the technology was so successful in helping customers obtain rate quotes that it is now used for scheduling pick-ups, tracking packages and ordering supplies. Another case where speech recognition has left its mark is with United Airlines. It was a huge test for the technology by this early adopter. But the business case was compelling The airline handles an estimated 1.5 million non-revenue producing calls a year from employees to list themselves for standby seats on United's 2,300 daily flights to 130 destinations. After a four-month development period, a speech recognition program was launched at the end of 1997. The system provided immediate service, 24 hours a day, seven days a week, and answered queries using either directed dialogue (telling the caller how to interact) or natural language ("understanding" information and requests however the caller chooses to deliver them). The result? The system had a transaction completion rate of 97 percent. It has been so efficient, that more than 70 percent of employees voluntarily choose to deal with the speech recognition system instead of waiting for a reservations agent. And it saves United millions of dollars a year. The employee application was so successful that United soon began work to implement four additional applications for the general public. Before speech technology, customers checking on the status of a certain airline flight were asked to press 1 for flight information; press 1 for arrivals or 2 for departures; press 1 if they knew the flight number, 2 if they didn't; key in the flight number and a pound sign; press 1 for today and 2 for tomorrow. If they didn't know the flight number, they then have to press a different key, and so on and so on. Customers using the speech recognition system have a far easier time. They only have to say, "I want to check an arrival flight for today. I don't have the flight number." Already, four rounds of Q&A are eliminated. As with most speech customer service applications, the prompts are conversational and easy for even a first-time user to negotiate. Despite its attractive price tag, ease of use and popularity among consumers, speech recognition is not a shoo-in for every DTMF application. For instance, applications with a very short start-up time or a very short life are not a good fit for speech recognition - until application development becomes faster and easier. This is the area that the major speech players are concentrating on today, developing tool kits and pre-bundled dialog modules to cut weeks off the application development cycle. But when the elements are right, companies can come to view speech technology as a major component in delivering their customer service needs. Previously seen as unreliable, speech technology is now affordable, reliable, proven and in demand. For companies willing to take the plunge, speech technology promises to be a rare breed - technology that enhances customer service with virtually none of the trade-offs that have made DTMF technology a modern day annoyance.
Paul Kowal is president of Kowal Associates, a leading customer relationship management consulting services company located in Boston, Mass.
SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues