Billions of dollars are being lost in translation – literally

Artificial intelligence is revolutionizing the way humans communicate across the globe, but accurately conveying ideas and meaning across cultural barriers is tricky. Every language is different, and these differences don’t always lend themselves to effective communication. For instance, the Persian word for ‘grief’ is closely associated with regret, but in a dialect of Dargwa, the same word is more connected to anxiety. In American business, ‘difficult’ annotates something as being a challenge but not impossible, while in a Japanese boardroom, the direct translation signals something is “out of the question.” And when a German announcement was auto-translated to English, a request to stop “panic buying” (“hamsterkäufe”) became a request to ‘stop buying hamsters.’ These miscommunications are awkward, embarrassing, sometimes distasteful, but for global businesses, they can cost more than just being ‘red in the face.’ AI could be a way to avoid them.

With the world rapidly globalizing, businesses are finding themselves in new regional markets. In America, top global brands have now expanded their literacy to an average of over 30 languages. New marketplaces mean new streams of profit, but these opportunities aren’t without their challenges, particularly communication. One study from the United Kingdom found that language barriers with Arabic-, Chinese-, French- and Spanish-speaking countries could be preventing a massive £19bn in annual exports. So why aren’t standard, literal translations able to rectify this problem? Well, humans don’t speak literally.

Since its inception, AI has excelled at literal translation: taking a sentence and finding the best, one-for-one approximation in another language, but this approach often fails to capture crucial emotional subtext and cultural nuance. Over the past decade, a new group of researchers began leading developments in the AI field with a different mindset. One of them is Dr. Miles Wen, the CEO and Co-Founder of Fano. His company develops AI software to better understand the subtleties of human speech and languages.

At Fano, their AI language models are trained with large data sets to develop a learning process for comprehension and generalization. Dr. Wen describes it as a classification problem—giving the AI enough information to compare and contrast data and come to its own conclusion on meaning.

So, how does this new approach work in practice? In Hong Kong, the company developed an AI customer service agent that uses Sentiment Analysis. In spoken conversations, it was able to accurately understand the frustrations and complaints of callers, even when people were switching between Cantonese and English mid-sentence, a common language practice in this region. For companies that implemented this technology, the voicebot was able to successfully handle a majority of routine inquiries and transactions, leading to a 90% reduction in misrouted calls and 50% reduction in complaint handling time. This meant customers were not only able to be heard, but understood.

Human psychology plays a huge role in consumer behavior, with emotional resonance superseding almost all other factors. In a study of 100 advertisements across 25 different brands measuring marketing effectiveness, ads that elicited an emotional response generated a 23% lift in sales compared to those that did not. When looking at the neurobiological bases of feeling understood, researchers found it was a social reward, reinforcing personal connections. Conversely, the potential pain of being misunderstood could lead to the avoidance of certain social interactions. And this is directly seen in consumer buying patterns around language accessibility, with 40% of people avoiding companies that don’t offer native language support.

Fano is exploring how AI can mend this disconnect in translation and reconnect people to the globalized world and economy. Humans want to interact with each other; even across cultures, it’s part of our nature as a species. Yet, misunderstandings—or “difficulties,” if you will—in language have created barriers for this to happen. AI is knocking them down and putting bridges in their place.

Click here to explore other angles with HSBC.