When I was about 8 years old I became overly obsessed with the idea of moving objects with my mind — just think of a mini Darth Vader version of myself. But, truth be told, no matter how hard I tried, or how long I stared, nothing would really move.
Today, we might not be moving objects with our minds or be able to time travel (yet) but, we can do a lot more than what we’d expect merely 20 years ago. We can ask Alexa to play our favourite guilty pleasure on Spotify, have a robot open a door, and even send a car into space.
And that’s the thing with technology. It keeps evolving in every field and customer support is no exception.
Who’d’ve guessed that today we’d be able to have realtime chat in customer support rather than having to wait 48 hours to hear back from an agent over the phone to solve a billing issue?
And if you look at the numbers, it’s quite impressive. According to this report by Forrester, the number of customers that use live chat over phone and email has increased by 50% since 2012. In fact, live chat is now so popular that it is the most preferred customer communication channel, according to this study.
No wonder then that more than two billion messages are exchanged between people and companies every month on Facebook Messenger alone…
Sure, realtime chat has been on the rise, and those of us who work in customer success have known this for years. But, with this trend also came the great responsibility of keeping things efficient and yet personal. It’s crucial for this kind of communication to mirror real conversations instead of just being awkward exchanges of information.
And this is where translation for customer service comes in. It’s just not enough to have multiple agents replying to customers if you can’t guarantee multilingual support. Or have a chatbot that only speaks English when a big part of your customers are scattered across the globe.
But, when most companies have already moved to a written customer experience and have been focusing a lot of their attention on realtime chat, can you still translate the messages in a timely manner? Is realtime chat translation possible?
Defining “realtime” and other limitations
By its very nature, chat demands a certain speediness. To understand if realtime chat translation is possible you need to look into the industry’s average. According to Zendesk’s report, support via messaging is most likely to be expected within 10 minutes and for realtime chat we’re talking about a first time reply of almost 2 minutes.
So, translation would have to come up on top of that. Let’s say an agent writes a reply. In theory, that piece of content would then be translated and sent back to the customer. But, how long does this translation take? And how does work? Is Machine Translation the only solution?
Anyone who has made any use of pure machine translation (Google Translate, et al), knows that it really is just not good enough for most but the most casual of purposes (for this and that reason and many others).
The truth is, translation is not as simple as it looks. It’s a difficult task even for humans, not to mention machines. So, with machines not accurate enough and humans unable to scale to the exponentially-increasing amounts of content, does that mean realtime translation is out of the question? Not necessarily.
How realtime chat translation works
At Unbabel we combine state-of-the-art Neural Machine Translation (NMT) with the efforts of tens of thousands of bilingual editors spread around the world who are able to deliver a fast translation with the best possible quality.
Imagine a customer reaches out to you by chat in Portuguese but you don’t speak that language. Unbabel automatically translates dozens of languages into your detected languages. With no language barrier in the way, you can instantly understand the query and provide a quick answer.
Once you hit send, our fully integrated translation pipeline ensures that the customer gets a reply in their language in a matter of minutes.
We might not yet be able to move objects with our minds, but translating conversations in realtime is here.