The 451 Group estimates just 5% of today’s IVR systems incorporate speech-recognition technology that lets users access information using spoken commands. The remaining 95% of the IVR market consists of much older, simple touch-tone menus from which users select choices by pushing buttons on a telephone, the firm says. According to most industry accounts, that ratio should be changing.
Many customers calling American Savings Bank to use its banking services find the automated attendant has something in common with them: a regional dialect.
The voice behind the tele-banking application is that of a local female, and the application’s expressions include “aloha” and “mahalo” to convey a familiar image, according to Renee Lum, assistant vice president and manager of American Savings Bank’s customer service center in Honolulu.
“The whole branding for our bank is based on our local image, and we wanted to carry that branding through our speech application,” she says.
Lum is scheduled to speak this week at the SpeechTek conference in San Francisco, where she will detail the bank’s decision to use technology from Intervoice for its migration from touch-tone to speech technology. The SpeechTek show is expected to draw 1,000 attendees to its educational sessions and 48-exhibitor exposition.
American Savings Bank deployed Intervoice’s development platform – the Omvia Voice Framework – and used the vendor’s professional services to design its custom banking application. Building a speech application is notoriously complex and requires multiple design tweaks and usability tests to make sure the application’s vocabulary and user interface are going to be well received by callers. Working with Intervoice, the bank brought in a range of customers of different ages, ethnic backgrounds and occupations to test the system’s usability before going live, Lum says. “We knew if it wasn’t easy, they were not going to use it,” she says.
American Savings Bank is among the early adopters of speech-recognition technology.
The 451 Group estimates just 5% of today’s interactive voice response (IVR) systems incorporate speech-recognition technology that lets users access information using spoken commands. The remaining 95% of the IVR market consists of much older, simple touch-tone menus from which users select choices by pushing buttons on a telephone, the firm says.
According to most industry accounts, that ratio should be changing. The emergence of standards such as VoiceXML – an XML variant designed so that content developed for the Web could interact with voice systems – is expected to drive speech-recognition technology into the mainstream. Indeed, vendors such as Nuance and ScanSoft are working to shift from expensive, proprietary platforms that require extensive professional services engagements to more packaged, standards-based technologies. At SpeechTek, Edify is expected to unveil a speech application tailored for consumer electronics companies.
But some industry watchers say the transition is stalled and the search for meaningful business returns remains elusive. It’s a vicious circle, says Steve Coplan, sector head for networks and media at The 451 Group. Vendors haven’t been able to build up enough demand for speech software, so they remain reliant on professional services. But the expense of specialized professional services in turn slows adoption, he says.
Coplan authored a research report released earlier this month that finds the speech industry won’t go mainstream until it becomes less expensive and easier to build speech applications. Among the factors stifling that transition are vendors’ dependence on professional services revenue, slow adoption of complex standards and packaged applications that fail to deliver the flexibility needed for broad deployments, he says.
Nonetheless, adoption of speech technology is growing, slowly but surely.
In 2003 and 2004, sales of touch-tone-based IVR systems declined, while speech-enabled system sales grew 15% to 20%, says Daniel Hong, voice business analyst at Datamonitor. Admittedly, the growth is not as strong as analysts expected a couple of years ago, he says. But IT spending is loosening up, and vendors are a bit smarter in mapping their products and services to the needs of customers. “Companies aren’t as skeptical as they were before,” Hong says.
Customers know the technology works, agrees Bill Meisel, president of TMA Associates, a speech consulting firm. While some of the first speech applications simply replicated the touch-tone applications they were replacing with speech commands, essentially giving callers the same list of options, today’s systems are much more sophisticated in what they can do, he says. For example, voice applications can tweak their instructions if a caller is recognized as a repeat customer who could take advantage of service shortcuts, Meisel says.
“Today, with speech recognition you can have a much more flexible conversation than you could with touch-tone, and people will accept that as an alternative to waiting for an agent,” Meisel says. “I don’t hear customers saying they’re afraid it won’t work. Now they’re saying it’s kind of expensive. That’s a very important transition.”
While today there’s a dependence on professional services – which drives up implementation costs – that will shrink as IT staff members learn the art of designing voice user interfaces and fine-tuning grammar. It’s a natural evolution of the technology, Meisel says.
Meanwhile, some companies turn to hosted IVR options to avoid the sticker shock of implementing speech technology in-house, Hong says. Vendors offering hosted voice applications and services include Angel.com, which will announce at SpeechTek the addition of call routing features to its service; BeVocal; Convergys and Tellme Networks.
