In their relentless drive to save overheads on call centres and other service facilities, bank chiefs are weighing up the many benefits of voice recognition software. Wendy Atkins reports.
In the past couple of years, debate has raged within the banking community about the best way to reduce call centre and back office costs without losing customers. One answer is to outsource these functions to countries such as India. Other banks have adopted strong marketing and advertising campaigns aimed at branding their organisations as high-quality institutions with high levels of customer service backed up by a human voice at the end of the phone or a friendly face at the counter.
New technology
While both methods have their merits, some organisations question whether they fulfil all of the cost and customer service requirements of many of today’s banks. As a result, there seems to be a renewed interest in a ‘third way’. Speech recognition is improving, and increasing numbers of consumers are now more favourably disposed to such technology.
This marks a shift from the mid-1990s when software developed in parts of the southern states of the US was said to be incapable of dealing with a European accent. Today, advocates of voice-activated, automated telephone lines argue that these can actually improve customer service: calls can be answered quickly and problems associated with call centre skills, staffing levels and operating banking systems 24/7 can be avoided.
One of the most recent to adopt speech recognition for banking functions is the UK-based Lloyds TSB Registrars, who now use the technology for administration activities associated with managing employee share schemes.
Fewer problems
Such environments are considered a natural home for automated procedures: they obviate the problems associated with lack of internet access; they provide a faster and more cost-effective alternative to paper responses via the post; and they solve the problem of live agents, who cannot be used for telephone call handling as there is a danger that some form of financial advice might inadvertently be given.
The system, known as the Telephone Application Service (TAS) system for employee share applications, was designed and is hosted by speech technologies provider SRC and uses the Nuance speech engine. SRC claims that call completion rates are high, and although some callers still call without the necessary documentation required to log in, speech recognition accuracy levels are about 90% and all details are confirmed with the caller before proceeding.
“On the relatively few occasions when there is a problem, it is still possible to fall back on a human agent to support the service,” reports Kenton Sanmogan, director of customer operations, SRC. “Fewer than 1% of the transactions handled actually requires any follow-up or human intervention.”
Getting it right Some speech-activated systems have run into problems with the pure recognition process. Dialogue design and conversational error-handling are important considerations in a good system. Integrators also have to grasp that answers can be confusing, so a second question may need to be asked to clear up ambiguities. Systems need to be designed to anticipate what the responses will be. And, because you can’t anticipate everything, human intervention needs to be available quickly as a back-up when there is confusion or complexity.
