Speech-recognition finds a home

SUNRISE, FLA. - ABN Amro Mortgage Group faced a big problem when home interest rates fell last year: Its call volumes went up. Way up. The lender found its call center swamped - with call loads jumping from a daily average of about 1,000 to about 5,000 in one day.

So the company deployed self-service voice technology and within weeks its agents doubled their productivity, according to Garth Graham, senior vice president in charge of e-commerce at ABN Amro Mortgage in Sunrise, Fla. When call volumes hit a peak, the application kept up with the human agents, processing as many applications as the agents alone could have normally. "So total production actually went up 300%," Graham says.

"It was impossible to put the number of agents that we needed in a call center to take the high capacity of phone calls you might receive based on changes in interest rates," Graham says. "So we built the application exactly like we would want an agent to handle things."

ABN Amro uses speech-recognition technology from SpeechWorks, which NetByTel delivers as a hosted service. NetByTel helped ABN Amro build a program that lets customers not only find out if refinancing would be a good idea, but also fill out mortgage applications over the phone.

"It ended up doing a lot more than we originally anticipated," he says.

ABN Amro has used the application for about a year, and Graham estimates revenue gains during that time total about $4 million. That's because the voice-recognition software to quickly screens callers, determining who should refinance. It then can take relevant information such as Social Security numbers and loan numbers and send the caller on to a live agent with the caller already well into the refinancing process. In some cases, the voice-recognition engine can complete the transaction without agent intervention. That means fewer dropped calls, less time on hold and higher agent productivity, Graham says.

ABN Amro's experience is not unique. Analysts say that businesses across the board can find revenue gains and savings thanks to voice-enabled, self-service applications such as those that SpeechWorks and NetByTel offer. The speech technology market - which The Kelsey Group says will increase from $505 million this year to $2 billion in 2006 - includes other players such as Nuance and Convergys Speech Solutions, as well as IBM and Philips, both of which have strong offerings in this area.

In a recent survey of SpeechWorks customers, The Kelsey Group found that customers saved on average about $1 million per year by using speech-recognition technology. Returns on the initial investment in the software typically were realized in less than one year.

Banking on voice A recent survey of SpeechWorks customers found that voice recog-nition software helped users in a variety of ways: • Saved an average of $1.02 million annually. • Achieved payback on their investment in about 9.5 months. • Had an average success rate with their automation programs of about 87%. • Reduced abandoned calls. SOURCE: THE KELSEY GROUP

The Kelsey Group conducted the study in the first quarter of this year, querying nine SpeechWorks customers that deploy the software in-house or access it via a service from Convergys Speech Solutions. The companies were in a variety of industries from healthcare and travel to financial services and telecommunications. The most recent deployment had been in operation for about five months, the longest for two years.

"We're pretty confident that there would be comparable economic performance from well-executed speech applications across the wider spectrum of offerings available," says Mark Plakias, senior vice president at The Kelsey Group. Similar offerings are available from Nuance, IBM and Phillips.

But Plakias notes companies that realized big gains took their time deploying the applications.

"Virtually every company we looked at had deployed one or two at the most applications in the past year. The pattern has been, let's prove this with an initial application and let's add on from there," he says.

The way companies use the technology varies, with some deploying it to upgrade existing touch-tone, self-service applications and others using it to automate processes that weren't automated previously. The way to avoid trouble is to determine where the application would work best.

"There is a learning curve," Plakias says. "But it will get shorter over time as these products get better-tuned and installations get more efficient."

RELATED LINKS