October 1999

Speech Recognition From A Buyer's Perspective

BY KEN ROKOFF, PRICEINTERACTIVE, INC.

You've finished reading up on the subject of speech recognition and now consider yourself the resident expert on selecting a vendor for your call center. You file the information on the shelf next to your reference book on how artificial intelligence is going to change your way of life in the coming years, and say to yourself, "I don't have time for this now."

Then, the phone rings. A local account executive from one of the better-known speech recognition vendors wants to "visit" with you. You acquiesce, hoping this technology can support your business objectives.

Sound familiar? This story highlights my experience with speech recognition and how I implemented this technology into a call center outsourcing environment. Here's how it happened, and how it might help those of you faced with the same challenge.

The account executive arrived at the appointed time and promptly began the time-honored task of attempting to harness the complexities of the LCD projector. After 15 minutes of fumbling, and with the assistance of our entire IT department, the projector was discarded and she put into place the generally expected backup plan. The representative quickly became alarmed as her acute senses identified me as one who will look ahead and ask questions prior to the presentation.

But I gave her the benefit of the doubt and allowed her to proceed for about five minutes before I searched for the coveted price page. Bingo! I discovered a chart resembling the most recent version of the U.S. defense budget. While speech recognition technology is less costly than it was in the past, it still is fairly expensive at between $500 and $1,500 per port, depending on the complexity of the application and number of modules required. The representative acknowledged my initial reaction to the pricing/figures and retrieved the ROI charts that headquarters had prepared for just this situation. (I was obviously not the first person with this reaction.) She quickly illustrated how this technology will reduce my call center and interactive voice response (IVR) costs through unbelievable efficiencies. Unfortunately, the various tables and payback calculations conjured up memories of the three-dimensional chess game once enjoyed by Captain Kirk and Mr. Spock.

I was revived without the need for smelling salts, and I quickly became a qualified prospect on my representative's weekly report. Three days later she called back to arrange for a follow-up meeting with her vice president of Business Development. So began my voyage into the amazing world of speech recognition technology.

While this story may be amusing, it reflects the experience of telecom professionals. Speech recognition technology is still the domain of early adopters. My colleagues in the financial and travel industries have blazed the trail as both innovators and early adopters. If history is any indicator, a chasm crossing to the early majority, or pragmatists, is still another 12 to 18 months away.

Nevertheless, speech recognition technology has arrived and will significantly impact the way we do business over the phone in the very near future. I will describe here the applications that can be greatly enhanced today through speech recognition. Most important, I will also detail how use of this technology by IVR and call center service providers will make this very useful technology an affordable reality in the coming months instead of years.

The Need
Let’s return to our story. My new friend, the account executive, arrived with the corporate vice president in tow. We entered the conference room and quickly agreed to a basic question-and-answer format.

In a moment of weakness, I threw a softball to my new comrade. “Why does anyone with a well-managed call center that already uses IVR service need speech recognition?” I saw the smile on his face long before the ball left the park. Here were his key points:

• Today’s consumers are demanding better, faster customer service and access to real-time transactions.
• Call center operations that continue to rely on live operators and menu-driven, touch-tone systems are missing an opportunity to increase customer satisfaction and save on costs and wait times.
• Call centers realize that balancing staff resources with unpredictable call loads is a constant challenge. Finding and retaining quality agents is time-consuming and expensive.
• Various routine requests come into the call center, leaving little time for selling. Customers are demanding anytime, anywhere access, superior convenience, immediate visibility to information, less time per transaction, “hands-free” operation and secure transaction handling. More and more, organizations that previously focused on cost and service are now expected to contribute directly to the sales effort.

The vice president then pointed out that a new solution for these challenges is automated speech recognition (ASR). He noted that ASR can:

• “Speech-enable” repetitive sales-support activities to respond to product information requests and completing product applications,
• Drive access to additional customer data and product scripts to directly support call center agents in the selling effort,
• Capture customer demographic and satisfaction data,
• Be harnessed to support the goals of other internal departments such as human resources and information systems and execute repetitive processes like benefits enrollment, job postings and system help desk support more cost-effectively,
• Cost-effectively recognize both simple commands and more complex phrases,
• Improve the ease and quality of the caller’s interactions with the system, and
• Capitalize and build upon the success of touch-tone applications, making it the next logical step in the evolution of self-service.

The Current State Of IVR In The Call Center
I realized after this compelling case that it was time to seriously consider implementing some speech recognition technology. I mentioned that I spoke with my IVR developers, who would love to adopt some of this technology. Their only reservation was finding an existing application that would benefit from ASR. I closed our second meeting by defining IVR as it applies to the call center universe today and suggested that my new friends return within a week with a detailed presentation on how they could help.

Here’s my take on how IVR has brought enormous efficiencies to the call center in four critical areas:

Call Management. The largest challenge in operating a call center is to resolve high-volume traffic with minimal call holding time. IVR systems are now commonplace in all major call center operations with the specific strategy of off-loading common functions such as frequently asked questions, account balances, shipping status, warranty registration, etc.

Information Capture. The largest challenge and expense associated with direct response advertising is the capture of leads and information requests on inbound toll-free lines. Live operator centers, extremely costly for these campaigns, are often unable to staff for the traffic generated by primetime media advertising. Such IVR-based systems can capture information more quickly and efficiently. But some of the more brand-conscious companies view IVR as lacking a human element.

Dealer Locator. This application front-ends the live operator to answer the common question, “Where is the nearest dealer location and how do I get there?” The call center typically uses automatic number identification (ANI) lookup or prompts the consumer for a ZIP code or phone number to determine the proximity of the nearest dealer(s) through a geo-coded lookup. The response can then be delivered through a live operator or, more efficiently, through an IVR response.

Commerce/Transaction Processing. Basic commerce applications such as Home Shopping Network’s Short Shopping program have begun to introduce IVR into the shopping/transaction processing realm. These programs are able to acquire and validate credit card transactions. However, they are limited in their capability to capture extensive information about the consumer without a pre-existing relationship.

Unfortunately, some of the most complex applications, including travel information and brokerage transactions, are difficult to implement through DTMF-based (dual-tone multifrequency) IVR systems. The very nature of these services is based on alpha and numeric key values (e.g., city codes, ticker symbols). Being laborious to enter through the phone, these types of applications became prime targets for speech recognition vendors.

Speech Recognition’s Early Adopters
During a subsequent flight between Washington, D.C. and Portland, I had the opportunity to educate myself on the trials and tribulations of the early speech recognition adopters. What I learned is rather compelling.

United Airlines was able to automate an employee reservation application, which supports nearly 1.5 million calls per year. Employees are able call a dedicated number, enter I.D. and PIN numbers and reserve their own travel without having to speak with a reservations agent. The total time to deploy the service was four months.

E*TRADE launched a speech-recognition-enabled service known as TELE*MASTER in December of 1997 that offered full transaction processing of stock, options and mutual funds. I tried a similar service on my own and discovered a level of intelligence beyond my expectations when my request for a current quote on “Big Blue” was answered immediately and accurately.

Bringing Speech Recognition To The Call Center
My account executive returned one week later, along with the vice president of Engineering, and continued our discussion on traditional IVR-based call center applications identified previously. Since the first meeting, one of my business partners had expressed an interest in a speech-enabled application, so I paid closer attention.The team reviewed the benefits of speech recognition:

Call Management. As mentioned earlier, call management is a key strength of IVR systems today. Unfortunately, the touch-tone user interface is limited in its flexibility and depth of interaction. For example, in a modern, technical-support help-desk environment, companies are trying to drive inquiries to the least expensive solution — first, to the company’s Web site, then to an automated system and finally to a live customer service technician, although it is hoped that this is an infrequent event. With automated speech recognition, we can resolve more caller inquiries by interfacing expert systems and large-vocabulary speech recognition to ensure a better trouble resolution solution — a solution that is not possible in a traditional touch-tone-based frequently asked question (FAQ) environment.

Management of calls into geographically dispersed multiple call center environments also benefits from automated speech recognition. Using ASR will allow call centers to quickly identify the specific skills required to handle the caller in the most efficient and least costly manner. For example, as a retailer with multiple call centers across the country and agents trained as experts in specific business functions, I want to identify the caller’s needs up front. I need to send the caller to customer service to help him or her, and obtain higher first-call-resolution rates, if I haven’t already resolved the caller’s needs in the automated system. Using ASR, I simply guide the caller to the appropriate customer service personnel. This is a vast improvement over the traditional touch-tone environment where I may force the caller to negotiate several levels of prompting, resulting in a lower level of customer satisfaction. ASR increases customer satisfaction because the caller spends less time on the telephone. It increases agents’ satisfaction because they can become experts in a specific knowledge area.

Information Capture. This classic IVR information capture application enables millions of transactions to be handled every month by both IVR systems and live operators. These programs tend to be high-volume, direct response advertising-driven programs that challenge live operator call centers. Automated speech recognition used in this environment can provide a user-friendly caller interface while driving out significant costs. ASR can be used during the information capture process to acquire name and address information, specific demographics and other information that would traditionally require a live operator or transcriptionist.

An example of information capture is a major brokerage house diving headfirst into the Internet brokerage business with a multimillion dollar, prime-time television advertising campaign. This application needs a high-touch, customer-friendly first encounter when interested parties call the toll-free number advertised on “Friends,” but the brokerage house wants the campaign to be cost-effective. ASR allows callers to request information from the brokerage house or even fill out an application and open an Internet brokerage account without involving a live operator, unless the caller specifically requires one. This process ensures that private financial information, like bank account numbers, credit card accounts, household income, Social Security numbers, etc., are divulged only to the brokerage house. The caller can start trading, confident that his or her personal information is secure. The brokerage has very cost-effectively created a new customer.

Dealer Locator. Dealer/store/outlet locator programs will benefit from ASR because potential customers can request location information without using telephone numbers or ZIP codes.

Where Do We Go From Here?
Commerce/transaction processing is the Holy Grail of voice-based automation applications. Through ASR, transactions that once could be completed only by speaking with an expensive live operator can be automated. Business-to-consumer, catalog purchasing, business-to-business transactions, inventory and shipping inquiries and complex alphanumeric interactions are all possible with ASR.

The success of several companies already using this technology has suggested that the public accepts automated transaction processing. So it makes sense economically to automate as many of these processes as possible. Using ASR, this application allows companies selling products from inventory with several thousand different part numbers or part names to automate the transactions and avoid the high expense and variability of live operators. Customers order what they want, when they want. They can review the shipping status of previous orders or their own account information on demand, 24 hours per day, whether or not the call center is open. This is true customer service.

Following is a shortlist of tasks to consider when bringing speech recognition to your call center.

• Identify a target project in one of the four IVR-based call center applications mentioned above. Some vendors have already built specific dialog modules that eliminate the need to construct large vocabularies for targeted call center applications.

• Evaluate your internal resources. Make an honest assessment of the capabilities and availability of internal resources to implement new technology. Be very careful to determine the compatibility of existing IVR infrastructure with available speech recognition technology so you don’t find your IVR system vendor with his upgrade order pad outside your office.

• Outsource. Unless you have a large IVR support staff, vast amounts of time to attend training and a significant budget for consultants, you are well advised to turn your project over to an experienced service agency. Several of the more prominent service agencies have partnered with technology providers to offer speech- recognition-enabled solutions at costs below that required to build an in-house infrastructure.

• Stop, look and listen. The rules you teach your children for crossing the street apply in the deployment of new technologies. Let your outsource provider, whether IVR or call center, teach you the ins and outs of this very beneficial technology.

• Last, enjoy the experience. Speech recognition is a high-visibility technology and, if implemented properly, will differentiate your company as customer-friendly and reduce cost.

Ken Rokoff is the vice president of sales and marketing of PriceInteractive, a provider of complex, high-capacity interactive applications.