杏吧原创

Eavesdropping call centre computers cut talk time

The next generation of call centre will utilise eavesdropping artificial intelligence to hunt down the information required while you chat

Phone a call centre and you are likely to spend ages on hold listening to canned music 鈥 and then find the operator cannot find the information you need. But an artificial intelligence system that hunts down the required information is aiming to slash the time people waste this way.

Using a mixture of speech recognition and search engine technology, the system, being developed by IBM, will trawl a call centre鈥檚 databanks for the information a customer wants and present it to the operator before the caller has finished explaining what they want. By giving operators rapid access to the right information, calls will be dealt with faster.

The system works by listening in to the conversation and identifying keywords spoken by the customer. It then flashes up the most relevant information it can find onto the operator鈥檚 screen. IBM project leader Johan Schuurmans says an early version of the system allowed calls to be handled about 20% faster than normal.

Call centres often take calls for several clients, perhaps an energy company, a bank and an insurance business. Each operator may have as many as 10,000 pages of product data to choose from. Koen Wolters, a computer scientist at the University of Twente in the Netherlands who helped develop the system, says finding the right page to match a customer鈥檚 query can be tough, and the caller frequently has to be put on hold while the operator searches for the relevant information.

Legal warnings

For instance, the AI system will recognise words such as 鈥渕ortgage鈥, 鈥渞epayment鈥 and 鈥渋nterest鈥 and enter them into a search engine dedicated to the call centre鈥檚 network. In this case it would retrieve those pages concerning options on interest rates for repayment mortgages.

The speech recognition system can also be used to remind the operator of any aspects of the product that they are legally required to warn the customer about 鈥 that not keeping up payments on a mortgage could lead to the loss of the customer鈥檚 home, for example. If the system does not 鈥渉ear鈥 the keywords of that warning the operator will receive a sharp on-screen reminder before the call ends.

The prototype version of the system can search for only a handful of phrases and works only on a PC. But the first commercial version, which will go on trial at a Dutch bank later in September, will be able to detect 1000 keywords and will operate on the centre鈥檚 network server.

However, the technology 鈥 if it proves itself in the Dutch trial 鈥 could also have a downside that consumers may find less welcome: faster access to sales information will give the call centres the ability to make a much harder sell.

In the same way that the system reminds the operator to warn the customer, it will also remind them to 鈥渦psell鈥 on behalf of the client 鈥 convincing the caller that there are better deals to be had by paying a little more. 鈥淚t will certainly give a better focus on upselling opportunities,鈥 agrees IBM. So watch out for the techno-assisted hard sell.

More from New 杏吧原创

Explore the latest news, articles and features