Translating real-time speech into customer contact is no longer a thing of the future. Facility contact center Transcom uses it for various international clients for different languages. Real-time speech translation makes the world – or at least the issue of the tight labor market – a lot smaller again.

 

When setting up customer service, companies that work internationally are faced with the choice of which languages ​​they want to use to serve their customers. 'Minor' languages ​​are relatively expensive. But the shortage on the labor market also contributes to the fact that 'common' languages ​​are becoming more expensive. Real-time translation – also of speech – now appears to work well: facility contact center Transcom is using it for more and more clients. Ziptone spoke with Lars van der Las, VP Client Service Manager EMEIA at Transcom.

Applications

“We are currently deploying automated real-time translation across various contact channels. We have the solution for this Parloa brought home. We started with e-mail, mainly because the time pressure of real-time is not there. Then we applied real-time translation in chat. And we are now also applying the technology to voice for a number of our customers.”

Van der Las explains how it works: “Think of the situation where a Swedish customer asks a question in their own language and a trained employee in Cairo gets the question real-time on the screen in English. The agent talks back in English, with the solution translating the speech back to Swedish text in real time and then converting it into Swedish speech using an artificial voice – so the Swedish customer hears it with minimal delay.”

“Customers do receive a notification from this financial service provider that an automated translation service is being used, with a trained employee on the other side of the line. For this financial services provider we work worldwide in twelve languages, the automated translation option is used for three languages. We have started a pilot for Dutch and the first results look good.”

In the fight against the labor shortage

Real-time speech translation is a solution for 'small' or expensive languages, says Van der Las. “This not only concerns small language areas, but also larger languages ​​that are simply difficult to obtain. Think of Germany, where the nearshore market is becoming increasingly tight. But this solution is also perfect for languages ​​where there is relatively little supply in terms of volumes: Swedish, Finnish, but also Dutch. In addition: for complex issues you would like to offer live customer contact. That used to go to the second or third line. With this technology, all employees can work in multiple languages. Of course there is a tipping point: if conversations become extremely complicated in terms of content, the question is whether it would be better to be helped by a native speaker.”

Transcom fits it real-time speech translation with various clients, including players from the travel industry, consumer electronics, telco and utilities. Remarkably enough, the financial services provider is at the forefront when it comes to usage, says Van der Las. “We see that there is no negative impact on the CSAT. We measure it after every interaction, which yields a response rate of about 50%.”
If you want to know how it works, you can view Transcom's demo.

 

AI helps here too

Some companies used to offer their customers the option of opting for customer service in English, for example, at a certain waiting time. This option is also offered by some Transcom clients for real-time speech translation. “We are still looking for the most optimal moment: at what waiting time are you going to offer this option? Our analyzes show that younger customers make that choice easier than slightly older customers. As mentioned, we measure the CSAT afterwards, and we listen back to conversations - to coach employees, but also to look at the quality of the translation. We also listen in live and during the introduction we let a native speaker listen in on conversations so that this employee can take over a conversation if problems arise. Transcripts of all conversations are of course available. We can have these 'checked' by our conversational analytics tools, which can also make suggestions for improvement.”

Clients without cold feet

The preconditions? For real-time speech translation, that means clients without cold feet, Van der Las laughs. “You have to dare. The translation tool must be able to handle irregularities in the client's pronunciation and convert it properly into English. More importantly, just like with regular multilingual support, employees simply have to be fully trained – including on cultural aspects that have to do with the country where the customer comes from.” When agents in the Philippines or Egypt suddenly start working in German, Swedish, French and Italian in addition to their own language and English, they must therefore be trained in multiple cultures. “Employees can see from which country the customer comes.”

This solution can be rolled out very quickly, says Van der Las, precisely because it is separate from the content and is only placed as a kind of layer over the spoken interaction. Ultimately, the solution can also be linked to a knowledge base, so that an additional substantive check can also be carried out. But for now, that's not what the focus is on.

“AI will soon be able to handle ninety percent of customer questions”

How long will we need a human intermediary to serve people in their own language? Van der Las: “Customers have recently become more self-reliant. Live customer contact has therefore become considerably more complex in recent years. The next phase is that AI will soon be able to handle ninety percent of customer questions.”

He expects that the total population of customer contact employees will have shrunk by a quarter in the next five years. Reducing the number of customer contact employees by using all kinds of technological solutions can be done quickly, he says, but there is also a price tag: you cannot (and do not want to) just fire people. He therefore foresees that this shrinkage in the field will mainly take place through natural attrition.

But, he says, there will always be a part for complex problems or a personal approach. “In eight to twelve minutes, you as an employee can very well build a personal relationship with the customer, AI is not going to succeed.” (Ziptone/Erik Bouwer)

Featured, Technology Transcom