Customer loyalty is a fragile thing and easily lost if customers are frequently put on hold or greeted by robotic voice prompts whenever they phone a company's call centre.

This problem has triggered the development of a new generation of call centres that use sophisticated technologies such as speech recognition and customer relationship management (CRM) software to give better service and manage customer interactions across a range of channels.

According to a recent survey by Avaya, the US equipment vendor, half of all callers to a call centre will hang up if they are kept on hold for 45 seconds or more. What is worse, 15 per cent of callers will not call back.

Interestingly, playing the wrong music can cause the customer to hang up sooner. Avaya gives the example of one mail order company which greeted customers by playing the pop song "Hanging on the Telephone" by Blondie.

Many businesses are seeking ways to make their call centres more responsive. But they face conflicting pressures to improve customer service while at the same time cutting the cost of handling transactions.

One way to cut transaction costs is to develop more cost-effective customer contact channels such as the internet. Another is to use sophisticated technologies to automate call centres and reduce the average time agents spend on each call.

Existing automated call centre technologies, such as interactive voice response (IVR) systems, frustrate callers and cannot give complex information or handle transactions.

"IVR is not a natural way for people to answer and has been brought into use purely because of the potential cost savings," says Kevin Lucas, senior analyst at AMR Research, who specialises in call centre technology.

Many companies believe speech recognition could provide a better alternative to the complex menus and robotic voice prompts of a traditional IVR system. Once plagued by poor performance and high costs, speech recognition technologies have improved dramatically in recent years.

Avaya has added natural language speech recognition (NLSR) to its IVR offering. Based on a speech engine from Nuance, a US speech technology company, Avaya's NLSR system is designed to understand customers speaking in their natural, conversational voice at normal pace.

Avaya claims NLSR allows call centres to automate many self-service transactions and so free call centre agents to spend time on complex transactions.

Using speech recognition in call centres might seem paradoxical, as customers who telephone call centres expect to speak to a friendly human voice.

However, many calls generate little or no value to a business. In these cases, companies want to encourage customers to use cheaper virtual channels, such as the website or e-mail, to communicate with the company.

Only in the last resort, are they meant to phone the call centre and when they do, the time the human agents spend on each call is strictly rationed.

This is an area where speech technology is already playing an important role. For example, telephone services such as directory inquiries or collect (reverse charge) calls have to be done by a human operator and are costly to provide. Many telephone operators are turning to speech recognition technology to help reduce the time their call centre agents spend on such calls.

In such a system, the user calls directory inquiries and a human agent answers the call. Once the caller has given the details of the desired subscriber, the rest of the dialogue between caller and agent is fairly easy to predict. There comes a point when the agent can drop out of the call, leaving a speech recognition system to listen to the final words of the caller.

Telia, the Swedish operator, has been using this "call completion" technology since 1994. It clams that it has cut the average time agents handling a directory enquiry call by an average of 10 seconds, generating savings of $6m a year.

"The business case for this type of application is just staggering," says Minesh Patel, technical director at Vocalis, the UK-based speech technology company, which supplied Telia's system.

The phone is no longer the only method of communicating with a customer. The growth of lower-cost channels has transformed call centres into "contact centres" that seek to manage customer enquiries irrespective of whether they arrive by phone, are sent by e-mail, or are typed into a web page.

Buoyed by the internet boom, many companies have invested heavily in technology to build multiple channels that encompass the internet, call centres, Wap phones and even interactive TV.

But customers have learned to their cost that there is often little integration between these different channels.

The one experience more frustrating than being put on hold by a contact centre, is to attempt unsuccessfully to perform a transaction on a website and to then have to call the contact centre and repeat the details over the phone because the agent has no record of the failed transaction.

"It is important that the people who handle phone calls also know about the customer's e-mails, letters and web transactions," says Mr Lucas of AMR Research.

Genesys, the US-based equipment maker, develops contact centre technology designed to overcome this problem. The system integrates multiple channels using a concept called "universal queue". It identifies the customer - by recognising the phone number or e-mail address, for example - and "pops" the customer's transaction history on the screen at the same time as the agent answers the customer's voice call, e-mail or web enquiry.

"It is all about enhancing the overall customer experience and providing the company with a 360 degree view of the customer," says Kevin Kew, vice president for worldwide marketing at Genesys.

Sophisticated CRM software has become an important component of a modern contact centre. Genesys, for example, has relationships with leading CRM software vendors such as SAP, Siebel and Onyx.

While CRM has a reputation as a costly technology, many businesses believe the benefits promised by call centres can only be won by using CRM. For example, if an agent knows that the customer he or she is speaking to has made three previous calls to enquire about a loan, then the agent should spend time pursuing the loan - even if the customer has called on this occasion about a balance enquiry.

The flipside of CRM is that it also allows businesses to identify their least valued customers. While "platinum" customers will immediately get put through to a human agent, lesser mortals might first have to explain their inquiry to a speech recognition system or, worse, listen once more to "Hanging on the Telephone".