How to balance tech, leadership & linguists for global success from Xtensos CEO Paul McManus

I have worked with the Xtensos team and Paul in some capacity since early this year. He has always been a source of inspiration for me personally, and I have even featured Paul’s quotes on my Linkedin.

The experience with Paul and his team (shout out to Mitja & Theo & everyone else!) prompted this interview. As someone who started my career as a freelance translator, the subject has always been dear to my heart.
And like I have said in the past, AI is inevitable in all industries. Still, I have always appreciated how Xtensos are “pioneering new career paths for translators in the AI era” to use humans to “oversee quality management and providing the cultural expertise that maintains translation authenticity.” This was even before Xtensos became a Leadsourcing user!

Having said that, I hope everyone enjoys this piece from Paul on when businesses should decide when to employ AI localization.
You can often find me in a Starbucks, tethered to my phone (“fully-secured” for the more Information-Security focused readers), working and managing calls while enjoying a coffee. Many coffee connoisseurs tell me that they prefer other brands but for me I found Starbucks back in the mid 1990s when I first started travelling across the United States on a regular basis for work.
Back then, coffee was still a bit of a novelty to a young Irish guy. In Ireland, we were raised throughout the 1970’s and 1980’s on Irish tea brands like Barrys and Lyons (all Irish people know exactly what I’m talking about!) and so coffee, and more importantly the idea of being able to go to a coffee store/shop and work for hours was very cool and exotic.
Anyhow back to business, I have written this article as if I were having a coffee with you (in Starbucks, of course!!), and AI-driven localisation is the main subject of our discussion.
First sip of coffee, and the first question arises: How does a company know it is ready for AI-driven localisation?
A very good question, to which I have formed a simple theory after listening carefully to different people from the industry over the last couple of years. I believe most companies asking this question should draw a triangle, with each angle representing one of the company’s perspectives:
- The Commercial perspective – Held by senior leaders including all the way up to C-level. This perspective involves “I don’t care what AI we use, but we have to grow our sales and cut costs so get it done now!”
- The Engineering perspective – Held by technology enthusiasts across the company. This involves “Of course we are trying out this latest model which is supposed to be 3X better than ChatGPT, how cool is that?”
- The Linguistic perspective – Held by language teams, linguists, terminologists and other language professionals who believe that “ The gold standard still has to be the human translator. Translation reflects our brand and so it has to be perfect and it must be checked (and checked, and checked again) by human linguists.”

Which is the most important perspective? Which angle needs to be larger to fit all the concepts and ideas?
- Is it the Commercial perspective? Are the C-level executives really ready for AI or just ignorant about it?
- Are the engineers ready for this? Are they prepared to introduce AI to transformtheir department or are they just constantly attracted to shiny new models when they get released?
- Do you have translators and linguists in your company? Are they rooting for AI solutions to succeed, or do they fear it and try to hold back the tide like King Canute?
One thing is certain: to know if a company is ready to integrate AI into their localisation process, there must be a genuine belief that AI can and will actually work. People might not be onboard from the start, but it is important to remember the ethos used at Amazon and other successful companies in situations like this – “Disagree, but commit”. Passive resistance will not work.
Conclusion: for a company to know that it is ready for AI to work, the first step is to identify and assess the power of the three perspectives and eliminate resistance (both active and passive).
Let’s not let the coffee get cold, and so, while sipping, here is the next question that might arise: what resources should a company have in place in order to integrate AI solutions smoothly?
A good question. Perhaps the easiest way to understand this is through a simple checklist. If you tick one or more of these, you are already well on the way to being ready for AI Localisation:
- You have a clear content management strategy and perhaps a content management system;
- You already have a clean data structure, separated by content type
- Your company has clear project ownership: regardless of whether this is in a centralised or decentralised structure
Whenever we, at XTENSOS, first approach a client who is looking for a localisation solution (to reduce costs, save time, and improve quality) we discuss not only what we can offer but, first and foremost, what can be leveraged already in terms of the their existing content management strategy and system.
If they have a clean data structure, separated by content type then this is ideal but if not then we can start with that and helps them to get this in place.
Finally, we look at their project ownership structure and put forward the best strategy to deliver success regardless of whether the structure is centralised or decentralised. We also assess the three previous perspectives mentioned earlier and we work with the company to help them decide which of these are the best fit for the customer to realise a successful implementation.
Now that the caffeine is kicking in and the brain has some fuel for continuing the conversation, it might be a good time to share with you the solutions that we offer.
At XTENSOS, we build proprietary high-performing multi-lingual Large Language Models (LLMs), which are customised with each customer’s own data to ensure precision, accuracy and reliability for unique use cases.
If the client has 10,000 good clean data segments of translated data already available, then we can implement our fully customised multilingual translation model for them right away. If they don’t have that size of a dataset available yet then our strategy will have two phases, Phase one will be from now until they have that dataset built and then Phase two begins a longer-term savings trajectory.
This approach automates operations and delivers savings immediately to bring the customer to a point where the traditional translation budget effectively disappears.
Our return on investment (ROI) model will generate savings of up to 80% by replacing traditional translation costs with a smaller, more predictable monthly technology subscription.
As the coffee is going down very well now, let me share some common hurdles that companies face during the transition to AI Localisation.
One of the biggest hurdles that companies face is a lack of trust in AI technology mainly due to the over-hyped nature of AI.
What I can share is that – many Generative AI LLMs are not ready for delivering enterprise level translations without Human-in-the-loop supervision. This is because many of them have been built with an over-focus on English and many of these models cannot generate outputs with the consistency that is required.
Our models were designed to be multilingual from the start, and the actual language combination is practically irrelevant. Incidentally at XTENSOS, we also have Human-in-the-loop supervision, but we place more focus on the inputs and processes than only on the outputs like most or our competitors.
Another challenge is that customer must take the time to understand the process complexity, such as API integrations etc. in some detail. Our team helps the customer to understand this complexity using straightforward language and with no sales BS.
Oops, it looks like we’ve run out of coffee (a Venti one, I know), but hopefully not ideas. To recap, remember these three key points if you are thinking about transitioning to an AI Localisation approach:
- First, don’t get overly-influenced by linguistic concerns alone: it’s crucial that the company fosters a culture free of resistance to AI adoption and all three perspectives (Commercial, Engineering and Lingusitic) must be in balance;
- Second, take a close look at the data and content that you currently have, then make thoughtful, long-term decisions that support your team’s future efficiency and success;
- and Third, aim to implement only the right AI technology that truly delivers consistent human-parity quality and eliminates traditional translation costsin a sustainable way.
I know in our virtual coffee chat you couldn’t see my name and details on the coffee cup, but if Starbucks got it right, it would read Paul McManus. You can find me on LinkedIn if and whenever you’d like another coffee.
