When speech-recognition technology made the scene about five years ago, systems came with their fair share of technical glitches. The technology today has since progressed, allowing call centers to see the potential benefits of increased efficiency, employee productivity, revenue and customer satisfaction, as well as the ability to capture customer data at the front end.
In fact, call centers are adopting the technology at such a rapid rate, that David Bradshaw, chief analyst of CRM at research and consulting firm Ovum in Boston, projects that by 2005, 50% of the 3.3 million interactive voice response (IVR) ports in call centers in the U.S. will carry speech-recognition technology. "It's difficult to get hard numbers, because the companies are private," Bradshaw says, "but I believe it's a $3 billion to $5 billion industry."
Making the grade
According to Mark Thompson, director of research at management-consulting firm, Pelorus Group, in Raritan, N.J., the progress stems from the development of more robust systems, as well as heightened levels of accuracy. The algorithms have improved over the last year so that computers better recognize accents, speed and background noise. Thompson notes that most speech-recognition technologies jumped from experiencing 80% average accuracy rates five years ago, to currently 96%.
The technologies also include richer capabilities, Bradshaw adds. With many beginning to offer text-to-speech and built-in intelligence pieces that enable increased real-time personalization, the range of telephony interactions are broadening.
Although there are no published statistics on the penetration of speech recognition in call centers, Thompson is certain that adoption rates are rising. He says most callers are not put off by natural language and appreciate the speed offered by speech recognition. "The technology has been proven to work," Thompson says. "As more call centers implement it, more customers will get used to it."
Proven ROI also has lured call centers to speech recognition. "If you pay an agent $30,000 a year, and have costs on top of that, bringing the total cost-per-agent to $45,000, you can substitute that agent with a speech-enabled IVR system at $4,000 per port. All told, the system potentially pays for that agent," Bradshaw explains. Thompson adds that the cost of handling a call from a live agent ranges from $1 to $15. With speech technology, the cost falls to between $.10 and $1 per call. "And the ROI typically is three to six months. So in six months, the call center will have earned its money back."
Rather than replace their agents, most call centers simply want to increase their productivity. Speech recognition takes agents away from routine tasks and allows them to answer more-involved calls.
Stating its case
Boston-based SpeechWorks and Menlo Park, Calif.-based Nuance, two leading speech-recognition companies, according to Bradshaw and Thompson, are experiencing rapid growth. Steve Chambers, VP of worldwide marketing at SpeechWorks, says the company has grown by 40% to 60% since its product-line debut three years ago. Sales have doubled from one year ago. Chambers attributes the success partly to the industry's increased visibility, awareness and proven return. "More people are using [speech-recognition] systems and getting comfortable with them."
Chambers says the company's customers report shorter call lengths and a decreased number of opt-outs. Continental Airlines, for instance, has experienced 55% fewer opt-outs since deploying SpeechWorks, he says. AirTran Airways has reduced call lengths by 50%; and after the first three months of Amtrak's mid-2001 deployment, the system answered 72% of the 2.8 million monthly calls on the train status line, up from only an average 42% serviced by its traditionally automated service. "The key benefit translates to cost savings," Chambers says.
Vince Zagorski, director of call-center telephony at Amtrak, says his company's automated touchtone system previously used an extensive set of menus, with each directing the caller to spell out the words on the keypad.
"When you're on a car phone, it's not cool to be looking at a keypad," he says. "Over the years, we've gotten a fair number of complaints from people about this, and that the service contained too much information. When we found out about speech-recognition technology, we thought we could benefit from people saying what they want, rather than pushing buttons to get there. The benefits are significant, because of the number of people staying in the self-service application. They're getting the information they're looking for," Zagorski says.
Although Amtrak has not yet conducted an ROI, Zagorski knows Amtrak definitely will see a monetary savings. The company is at the tail end of testing a speech-recognition component for its reservation and scheduling line, and plans to release it nationwide some time in April. "If we can improve our metrics by 30%, that means that many more calls don't get answered by agents," he explains. "We like to have the agents working on revenue calls. We don't need to be taking up their time with calls such as finding out when a train is arriving."
Bradshaw projects that, by the end of this year, 900,000 IVR ports in call centers will have speech recognition. In response, the industry will work on increasing real-time personalization and perfecting the way the technology adapts to customers' behavior for improved accuracy rates. The industry will release a greater number of text-to-speech components, and vendors will work on a potential set of standards that can access both voice XML and Speech Application Language Tags (SALT) tools for more widespread use of capabilities.
To read more articles like this one, visit Peppers and Rogers Group's Web site at www.1to1.com.
All materials copyright 2002 Peppers and Rogers Group - 1:1 Marketing.