Communicating your message in multiple languages

By Linda Bicho-Vachon

Linda Bicho-Vachon looks at AI translation tools and whether they can reduce our reliance on human translators.

Four hanging speech bubbles on a blue background display "Hello," "Hola," "Nǐ Hǎo," and "Bonjour" in green, pink, yellow, and blue.

If you’re a multi-national company, communicating a message that conveys a certain meaning can be challenging to deliver in different languages. A poorly translated slogan can disconnect your company from your audience and be damaging to your brand. High quality translations are critical to ensuring that communications capture the intended sentiment of your message in multiple languages. So, is artificial intelligence (AI) ready to take on the role of translator?

AI translation tools have evolved significantly over the past few decades. The earliest machine-based translation technology relied heavily on a rules-based system that offered sentence-by-sentence translations but sounded awkward to the native language speaker. Today, AI translation systems such as Google Translate and DeepL are based on a neural machine-based system which translates text that sounds more natural and fluent.

At Microsoft’s Ignite 2024 conference in October, the software company announced the Interpreter Agent as part of its Copilot suite of automated services. This feature promises “real-time speech-to-speech interpretation during meetings”, and claims users will be able to sound like themselves, but in a different language. In 2024, Google Translate announced the addition of 110 new languages, proving AI technology continues to evolve.

Despite all these technological advances, one expert thinks companies shouldn’t forego the cost of hiring a professional translator, especially for its most critical communication needs.

“AI is not quite there yet,” according to Rosanne Lafontaine, a bilingual proofreader with over 25 years of experience.

While Lafontaine acknowledges the advancement of AI in fulfilling day-to-day translation needs, she says these technologies fall short in other areas.

“AI can translate words and recognize common expressions, but it can’t recognize context such as cultural or other nuances,” she says.

Lafontaine also says AI can make the mistake of translating words that shouldn’t be translated. She gives examples such as company names, job titles and acronyms to form words.

“AI also can’t recognize words with double meanings like “may” for the month or the verb,” says Lafontaine.

Mistakes can be costly

While hiring a professional translator may seem costly, communicating the wrong message can cost even more. Consider HSBC Bank which spent millions on a global marketing campaign based on their existing tag line, “Assume Nothing”, only to find in other countries, this phrase translated as “Do Nothing”. The bank spent over $10 million dollars to adjust their campaign’s tag line to a more universally accepted, “The World’s Private Bank”.

Frequently, a marketing phrase that is effective in one language, doesn’t evoke the same tone or sentiment when it’s translated word for word to another language. In some cases, an entirely different communication plan is needed to create a message that will resonate with consumers in that market. This process is referred to as transcreation.

“Just because all the words are in the requested language doesn’t mean they are right,” says Lafontaine.

Consider Nike’s slogan “Just Do It”. While it was very popular in English speaking countries, it lost its meaning when translated to other languages. To promote the brand in China around this phrase, the company created messaging in Chinese to explain the meaning behind the iconic slogan.

So, while AI can provide accurate and fluent translations, due diligence is still necessary to understand the market where your message will appear.

I thought back to my experience working with our internal communications team on a tag line to promote the release of our total compensation statements. In English, we settled on “Your salary is just one piece of the pie,” to highlight to employees the value of the company’s benefits and perks. Our graphics team went on to create images to align to this promotion.

When we sent the materials to our Quebec colleagues, they advised that this tag line would not work in French. They suggested a more suitable phrase, which when translated into English said, “Your salary is just one piece of the cake”. Since we were so close to the launch of the statements, we settled on the more appropriate French phrase for our french-speaking employees. We asked our graphics team to produce a second set of images to align to the French message.

The human touch

What about Microsoft’s Chief Marketing Officer’s claim that their Interpreter Agent can make you “sound like you in other languages”? Lafontaine says AI tools are limited in their ability to make someone sound like a human in other languages.

“In a video script translation, AI can’t pick up on differences implied by the inflection or emphasis on certain words in actual video,” says Lafontaine.

Tips for selecting a translator

Once your business is ready to hire a translator, here are some dos and don’ts for making the most of this relationship.

Do’s

Pick a translator who is a native speaker to the language and who has a degree in translation.
Select a translator who is specialized in the industry they are translating materials. This can be especially critical in the translation of legal and medical documents that contain specific terminology.

“A translator who is not only an expert in translation, but specialized in the field in which they are translating is important,” says Lafontaine.

Create a database of words and expressions in another language that are unique to your business.
Bring in a translator during the early stages of your communications planning.

Don’ts

Don’t bring in a translator who is not fluent in the desired language.
Don’t have a translator review work that has already been translated by AI.

“The work involved in revising AI-translated material always takes longer than if a translator had started from scratch,” says Lafontaine.

Don’t publish AI generated translations until they’ve been reviewed by an experienced person in the field.

About the Author

Linda Bicho-Vachon is a journalist, writer and editor with experience in storytelling for corporations and media organizations. As a chase producer for one of Canada’s national news outlets, she produced informative segments on the day’s top stories covering a variety of topics including politics, business, health, and entertainment. Her written work has appeared in BNN Bloomberg and Benefits Canada. Prior to this, Linda spent 20 years as a human resources manager where she led dynamic teams, drove strategy, and produced employee communications designed to engage employees and boost brand favourability. Whether it’s a news story, feature, essay or business communications, Linda brings a unique perspective that will make your story stand out. You can connect with Linda through LinkedIn.

Communicating your message in multiple languages

Return to the January 2025 Issue of Communicator

READ MORE

Communicator January 2025: Accessibility and Artificial Intelligence (AI)

10 Goals for Growth for 2025

How AI Can Make us Better Professionals