How Mother Tongue AI is Opening Up the World of Transcreation

As any linguist will tell you, expressing ideas is a surprisingly creative exercise. For Omnicom’s transcreation unit Mother Tongue, human creativity has always been at its core. And its new artificial intelligence offering, Mother Tongue AI, is no different. This is AI with a human touch.

Mother Tongue AI will be a discrete part of the business and it aims to 'support global brands in search of innovative ways to enhance the human-AI experience for their users worldwide'. It will be headquartered in San Francisco and run by global lead Valeria Barbero, who is backed by client director, Rhianna Floresca.

Mother Tongue's CEO Chris Hodgson Castillo says, “We are really proud of our work improving high-profile LLMs over the last 18 months. Through this dedicated new unit, we can now take that expertise and experience to clients in the tech space and beyond.” 

Mother Tongue AI has been a natural evolution for the company. Founded in 1990, they joined Omnicom in 2012, and work across Omnicom’s agencies to help with creative translation of advertising materials. Over the past 8 years, however, they have increasingly worked directly with blue-chip brands, predominantly in the tech industry, on their continuous localization needs. Mother Tongue AI was a natural continuation of this work, as Floortje Foster-van der Laken, Mother Tongue’s Business Development Director, explains.  

“Mother Tongue has been traditionally rooted in the advertising industry, mainly working with ad agencies to adapt creative campaigns. It’s when we started applying those same creative principles to UI/UX and longer technological content, working direct-to-brand, that presented the opportunity for Mother Tongue AI and Mother Tongue’s next step into the tech world.”

Mother Tongue AI’s solutions are segmented into four main categories – Conversational UX Optimization, Content Management, Bot Persona Development, and Trust & Safety options.

As much as it’s a cutting-edge tech service, it’s also one that has a very human touch. Flo explains that human feedback is key to optimising these models.

“Reinforcement Learning from Human Feedback (RLHF) is at the core of our Content Management solution. Essentially, it’s about having AI-experienced, trained linguists provide rewrites for and feedback on the output of an LLM. This feedback is then used to optimize future generative output of the LLM, thus improving the quality of the model.” 

Although Mother Tongue AI is specifically dedicated to improving the human-AI experience, and improving the quality of Generative AI language models, they are currently not using GenAI for traditional transcreation of ATL campaigns.

“Our approach to transcreation of ATL campaigns hasn’t changed, and is still performed by advertising-agency experienced, market-based copywriters. Although we are utilizing and continuously exploring machine solutions (AI, but also NMT, Neural Machine Translation) for our long-form translation, which is less creative in nature.”

Mother Tongue AI has just officially launched but it’s something that they’ve been working on with some existing clients already. It’s taken 18 months to develop and, even as it launches, is still evolving and improving - as is the way with artificial intelligence in 2024. 

“We believe AI is here to stay. It's going to be about being part of that story, and how we can actually make AI a better human experience,” says Floortje.