27/08/25

Few technologies have changed the translation industry as fundamentally in recent years as artificial intelligence. oneword has been working with AI systems for over eight years – a period of intensive innovation and associated experience. One insight has clearly emerged from this: AI can speed up processes enormously, but excellent quality can only be achieved in conjunction with the expertise of experienced translators.
In an interview with Sara Cantaro, Head of Marketing Management, oneword director Andrea Modersohn explains exactly how companies utilise this added value – and what leading role oneword plays in this.

Sara Cantaro (SC): Hello Andrea, thanks for taking the time to explore the topic of AI from your perspective.

Andrea Modersohn (AM): It’s my pleasure. After all, it’s a topic that is extremely important for our industry.

SC: That’s true, of course. Some companies think that AI can handle translation processes completely without human intervention. What do you say to that?

AM: Systems like ChatGPT impress us all. And they are also often used for translation. What works for e-mails, presentations or summarising texts gives the impression that these systems work perfectly or at least almost perfectly. However, when it comes to professional translation, for instance in technical documentation, these systems quickly reach their limits. Because it’s not just about language, but also about the level of risk.

Let me give you a somewhat drastic example: nobody would leave the translation of software for controlling a nuclear power plant to an AI alone. The reason is simple: AI can do a lot, but it makes mistakes. And this is precisely why human expertise remains indispensable, especially where precision, context and responsibility are required.

“And this is precisely why human expertise remains indispensable, especially where precision, context and responsibility are required.”

Andrea Modersohn, Managing Director oneword

SC: I understand. What advice would you give to companies that believe they can manage completely without human translators thanks to AI?

AM: The dream of fully automated translation has been around for decades – and it will probably remain a dream for quite some time. (laughs) As a linguist, I say: Language is a highly complex system, and then there is the demanding content that we translate every day. Human checking is simply indispensable for high-quality results. In many scenarios, the use of AI alone simply makes no sense.

SC: Where is the expertise of translators indispensable? Can you give us an example?

AM: Large language models can deal with language, but they lack the human ability to think, interpret and also fill in missing information. Furthermore, the current models are neither specifically trained for translation nor customised for specific industries or organisations. And they have difficulties interpreting ambiguities. Many source texts also have weaknesses that first need to be recognised and resolved. In all these cases, the work of translators is indispensable. They not only contribute language and specialist knowledge and their culture, but also take on roles as quality managers, consultants and language engineers. Their particular strength lies in combining a deep understanding of language with technological expertise.

SC: How has AI changed the work of translators?

AM: Translators used to do the translation themselves, but nowadays it’s often done by AI. Translators check and adapt the translation – this is a completely different process and requires completely different skills.

SC: What problems arise in the process?

AM: Checking and correcting, in particular, are demanding tasks. AI texts appear fluent at first glance, but their content often does not match the source text. For technical texts, the average change rate is still between 50% and 66%. In other words, mistakes creep in that a translator would not have made themselves. This is a real risk – you have to work extremely carefully.

SC: I can imagine that very well. How does oneword ensure that such errors are avoided?

AM: Our approach is very similar to the one we use for traditional translation, but with a few differences. From the very beginning, we rely on highly qualified specialists, a multi-stage quality system and – crucially – core translators. This means that our clients work with the same post-editing teams over the long term, so the results remain uniformly accurate and consistent. In addition, we use extensive feedback analysis to continuously check and improve the quality of the output. In some respects, this makes the process even more complex than with traditional translation.

SC: When does AI-supported translation make sense – and when does it not?

AM: This depends heavily on the subject area, the language combination, the type of text and also the requirements. We do not exclude any text from the outset or categorically, but always check individually whether machine pre-translation is suitable. But if too many corrections and improvements have to be made, at some point MTPE will no longer be worthwhile in purely economic terms. Then a traditional translation is the better choice.

SC: It feels like something changes every week when it comes to AI. How does oneword stay up to date?

AM: We have been working with AI for a long time – specifically since 2017, when the first neural systems such as DeepL appeared. And since 2019, our MTPE processes have been certified in accordance with ISO 18587. This gives us a clear advantage, even though LLMs have of course changed a lot in recent years.

SC: So this really isn’t such a new topic for oneword?

AM: Exactly. The post-editing process has essentially remained the same, which means we can provide our translators with good support and targeted training. We also have our own MTPE team, who are continually working on all developments relating to MT and AI, expanding our products and continuously reviewing and optimising our processes.

SC: Finally, I would like to take a look into the crystal ball. Where will the translation industry be in five years’ time and what role will oneword play in it?

AM: Good question – next question! (laughs) In all seriousness, nobody can say today exactly where we will be in five years’ time. What we can say with certainty, however, is that we at oneword do not wait and see, but set trends ourselves. We develop solutions that combine the best of both worlds: state-of-the-art AI embedded in specialist human expertise.
We are currently investing in our own LLMs and deploy software that already enables us to provide our customers with the best possible results. We work in a very customer-centred way – because for us the crucial factor is what solutions our customers need. And these are precisely what we implement, flexibly and without the hindrance of technical blinkers.

SC: Great closing words! Thank you very much for the interesting interview, Andrea.

“We develop solutions that combine the best of both worlds: state-of-the-art AI embedded in specialist human expertise.”

Andrea Modersohn, Managing Director oneword