Computers 'destined to replace call-centre staff'

By CHRIS BARTON

Paul Magee believes Government plans to promote New Zealand as an ideal location for call centres by offering subsidies or incentives are ill-founded.

Mr McGee, managing director of VeCommerce, says his belief is not just because of the cost of setting up buildings to house armies of shift workers around the clock, but because in a few years such centres will not be needed.

He believes the people doing repetitive call centre tasks will be replaced by computers.

Specifically, he is talking about computers that have natural language speech recognition - something VeCommerce is promoting both here and in Australia.

The most recent example of what the new technology can do is seen in the $A3.5 million ($4.4 million) system at Queensland's TAB, which allows punters to call up and place their horse racing bets through voice commands to a computer.

So successful is the system, that all but one of Australia's TABs are now planning to use it.

VeCommerce, formerly Scitec, is listed on the Australian stock exchange and has a 26-strong research and development team in New Zealand, thanks to the $3 million acquisition of local software house Voice Link 2sfr1/2 years ago.

Customers in New Zealand, including the New Zealand TAB and Mobil, currently use VeComerce's older touch-tone phone technology to interact with a computer by selecting services from a range of menu choices.Tranz Rail is the first company in New Zealand to move to the next stage. It is currently working with VeCommerce to develop a natural language speech recognition system to tell customers about train schedules and enable them to buy tickets.

The system is expected to be available early next year.

Mr Magee advances two arguments on why computers running natural language voice command systems are better than call centres with real people at the other end of the line.

The first relates to the problem of high staff turnover - that answering mundane, repetitive customer inquiries is not a job that many people like to do for long.

The result, according to Mr Magee, is that call centres are constantly training new staff and that mistakes occur, due either to the onset of boredom in experienced operators or the inexperience of new trainees.

"You have to accept the reality that human beings are not very good at doing routine transactions really well," said Mr Magee.

A similar effect occurs at the other end of the line with touch-tone interactive voice response systems. Give a customer more than three choices from a menu and mistakes start to happen - creating customer frustration.

But it is the second argument - cost - that is the most compelling.

The cost of a call centre employee answering about 25,000 calls a year, says Mr Magee, is about $50,000 a year.

The cost of a port on a natural language system that can answer about 100,000 calls a year is $25,000.

But Mr Magee does not claim natural language systems can do everything.

He says it is a balance of "high-tech and high-touch" and the systems have to be used in conjunction with call centres manned by people to deal with the more complicated type of customer inquiries.

So just how accurate are the natural language systems?

Mr Magee says the Queensland TAB is getting about a 90 per cent hit rate in placing bets by voice response to computer prompts.

To cope with Australia's rich ethnic mix, the language model was developed from voice recordings of a sample of existing TAB punters.

The system is also smart enough to know what type of phone is calling and to adapt to effects like background noise.

But Mr Magee says the real ingenuity is not in the speech recognition engine that converts phonemes (a language's distinguishing units of sound) to the electrical signals the computer understands

but in the prompt script the computer follows when unexpected events occur - like a cough, or when a punter says hello to someone while in the middle of placing a bet.