chloeveltman.com > features

Call centre technology: Voice recognition could be sound investment
18 September 2002

The Grand National has always been one of the busiest days of the year for Britain's bookmakers, but this year, relief was at hand. For the first time in the history of the big race, two bookies, Tote and Littlewoods, used a voice recognition system allowing customers to place bets over the phone without needing to speak to a live agent.

The system was implemented by SRC (The Speech Recognition Company), a UK-based speech recognition solution provider, using a speech recognition engine developed by California-based Nuance.

Normally, 700 extra call centre staff would have been drafted in for the day to manage the extra volume, but according to Nuance, this year the bookies hired no extra staff and service levels of calls answered within 20 seconds increased from 60 per cent to around 90 per cent.

"Call waiting times were slashed because the speech recognition system could handle thousands of callers at a time and the bookmakers received fewer complaints from punters about incorrect bets because the system was extremely accurate," says Kenton Sanmogan, head of solutions, marketing and consulting at SRC.

Business on Grand National day was so good that Littlewoods plans to use the system to handle day-to-day horse racing and football betting in future. Blue Square, the UK-based interactive betting company, also used voice recognition from Fluency Voice Technology as back-up on Grand National day.

The speech recognition industry has experienced peaks and troughs in the 30 years since its inception, as a result of the rate of technological progress and the hype cycle. Despite a general climate of disillusionment following the internet crash in 2000, speech recognition technology has progressed greatly since the early days, with increased R&D in the past five years fuelling faster innovation.

While historically, systems could only handle a limited vocabulary with speakers uttering one word at a time and pausing between each word, today's technology can handle full sentences, provides noise reduction, can compensate for different accents and dialects and requires less processing power.

The development of software standards and industry partnerships has also helped the sector to grow. The latest speech recognition systems can process a variety of customer calls, from call routing and information enquiries to transactions and basic technical help, without asking callers to navigate an endless maze of touchtone menu options. For callers who do not have access to a touchtone phone, need information in a "hands-free" environment, such as when driving, or don't want to wait on hold for a live agent, speech recognition is a powerful tool.

For companies, investing in speech recognition systems can make economic sense. According to Microsoft data, an agent-handled call costs anything between $5 and $12, whereas calls processed via a speech recognition system cost just 10 cents. Meanwhile, demand for call centres continues to increase at a rate of 2-3 per cent annually, and both staff turnover and labour costs as a percentage of the total remain high. Clearly, reducing staff numbers and freeing up agents to deal with more complex calls saves money.

"Successful speech implementations will reduce labour requirements in the call centre and decrease the closing time necessary for each call," says Gregory Neichin, author of a recent report on call centres by SRI Consulting Business Intelligence.

With such tangible advantages, speech recognition systems have begun to achieve widespread penetration in call centres across Europe and the US in the past year. From banks and florists to taxi companies and train ticket vendors, companies of all kinds are beginning to realise the economic and customer service benefits of fully or partially automating their call centres.

Today, according to consultants Blue Sky Consulting, there are an estimated 10,500 call centres in the UK, employing, according to estimates by the Trades Union Congress (TUC), more than 400,000 people. This represents about 1.7 per cent of the UK workforce. In Western Europe as a whole, the percentage of the total workforce employed in call centres is about 1.3 per cent.

Meanwhile in the US, research by business information resource Datamonitor reveals that 3 per cent of the nation's workforce are employed by call centres.Worldwide spending on voice recognition will reach $41bn by 2005, according to market research firm, The Kelsey Group.

Yet despite the economic benefits and recent advances in technology, the speech recognition sector faces significant challenges. One cause for concern within the industry is the current debate between two sets of technological standards: Speech Application Language Tags (SALT) and VoiceXML.

VoiceXML is the original standard - most of today's leading voice portals have developed their capabilities using this standard. The VoiceXML Forum, which was founded in 1999 by such companies as IBM and Nuance to develop the standard, currently includes more than 550 member companies. The SALT Forum, founded in 2001 by a powerful consortium including Microsoft, Cisco and Intel, is seen as a rival to VoiceXML by some industry insiders.

The main difference between SALT and VoiceXML is the fact that VoiceXML is its own mark-up language while SALT is superimposed on standard HTML, xHTML and XML pages. SALT focuses on developing "multimodal" applications (where, for example, users can input speech or text and receive text, graphic or speech output), whereas VoiceXML is based on telephonic applications.

While insisting that their standard is complementary to VoiceXML, the SALT camp also claims that VoiceXML is too rigid and does not integrate with existing web technology as smoothly. "Considering the development of call centre applications, it is clear that the objectives overlap, forcing businesses to choose between the two standards," says Chris Mulligan, a PA Consulting Group consultant.

Yet the debate shows few signs of becoming a full-scale war. Companies such as Convergys, the US call-centre operator and billing systems vendor, cheerfully support VoiceXML applications while at the same time belonging to the SALT Forum. Other companies such as Unisys, the big US IT services company, see an eventual convergence of the two standards.

"Unisys predicts an amicable marriage between the two standards that will drive speech technology even faster into the call centre environment," says Bill Scholz, Unisys's architect director of voice and business mobilisation solutions.

Ultimately, technology is less likely to hold the adoption of speech recognition systems back as much as human elements like branding and customer acceptance. With increasing numbers of calls being answered by automated voices, the human touch is as important as ever.

"The challenge for those using this technology is to decide which queries can be handled automatically and still keep the customer happy, and which can't," says Chris Harris, managing director of telephony provider Inter-Tel. "After all, the cost-cutting achieved through voice recognition will be counterproductive if customers are then driven away by unsatisfactory interactions with a voice recognition system."