By Ryan Falkenberg, 19 February 2024 – Originally published by IT Web Africa.
Here’s a challenge: phone the contact centre for your bank or insurance provider and see how many language options you get. In South Africa, besides English, you’ll probably get one or two at most. And while those options cover most of the population, many others will be forced to converse in their second or third language.
Now, try and repeat that experiment with the chatbot on that bank or insurer’s website. Chances are, there’ll be even fewer options. And when there are alternative options, there’s a fair likelihood that the answers in those alternative languages won’t be as good as they are in English.
There’s a good reason for that too. Until relatively recently, anyone wanting to build a chatbot or virtual agent capable of operating in multiple languages would have had their work cut out for them. You almost had to tell the bot exactly what to say. So, if the original logic was in English and you wanted the bot to also be able to answer in French, you had to write the logic all over again in French. It was a time-consuming, expensive, and extremely unscalable process.
Things improved as online translation tools like Google Translate came to prominence. These tools could translate your website content without you having to have different versions. The accuracy of these translations was less than perfect, however. And for companies like banks, this could be seriously problematic. The last thing you want is your customers in uproar because of an inaccurate and potentially offensive translation. Even if you didn’t have to duplicate the bot’s logic, these accuracy issues meant that organisations would have been reluctant to incorporate these translation tools into their chatbot or virtual agent offerings.
With the rapid advances in conversational AI capabilities, and more recently the acceleration in generative AI, the accuracy of language translation has improved significantly. So much so that we can train a virtual agent in English and, by leveraging advanced language translation services, have it engage with customers in many other languages across multiple channels.
As more and more organisations make use of the powerful translation services offered by Microsoft, AWS, IBM, Google and the like, the accuracy of their translations will keep improving. While the translation of certain conversations may still not be perfect, watch this space. Within the next 12 – 18 months you will see an exponential improvement in accuracy. Even if there is a human agent working with a virtual agent, they’ll be able to help someone who’s speaking a completely different language.
The benefits offered by virtual agents with this kind of translation capability in a country with as many official languages as South Africa are obvious. Organisations can offer support in most of the official languages without having to hire people proficient in each of those languages, or spend large amounts of time and money building and maintaining multilingual virtual agents.
But the benefits go far beyond that too. Almost any business with international aspirations will be able to provide useful, accurate, and relevant support to their customers, no matter where in the world they might be from. It’s also worth noting that these translation abilities won’t just be restricted to text-based conversations but voice-based ones too.
Make no mistake, these advances are coming at pace, especially now that the tech giants are pouring billions of dollars into generative AI. Businesses can get ahead of the curve by developing a contact centre strategy that includes multilingual virtual agents automating high volumes of their rule-bound sales, service and support conversations across their digital channels.