Chartered Institute
of Linguists

The skills AI has made more valuable, not less

 

 

When the European Master’s in Translation (EMT) network published its updated competence framework in October 2026, the headline was, predictably, about AI. The framework “reflects how AI is changing translation” and sets out what universities should be teaching future translators about large language models, output assessment, and “informed decisions.”

Tamer Osman notes however that perhaps more interesting are the competencies that remain which are firmly human: intercultural mediation, transcreation, data curation, and quality judgement, and argues that these are the areas where the profession is consolidating its value.

 

What emerges from the latest EMT framework is not a list of new additions to the translator’s toolkit, but a set of skills that are becoming more valuable, precisely because routine text production is being increasingly absorbed by tools.

For me the big question worth asking is not whether AI can translate. It is which human competencies become more valuable when it can.

 

Intercultural mediation: the skill of noticing what is not said


Intercultural mediation is the competence most easily romanticised and most easily misunderstood. It is not simply knowing that a gesture means something different in another country, or that a colour carries a different connotation. In practice, it is the disciplined work of noticing where a text assumes a shared frame of reference that the target reader does not have, and deciding what to do about it.

Consider a set of internal communications for a company expanding into a new market. The source text refers repeatedly to “the usual approval chain”.

For the source audience, that phrase carries a whole architecture of assumptions: who approves what, in what order, with what level of escalation, and what happens when someone is on leave.

A machine translation might render it fluently as ‘the usual approval chain' in the target language. The output would be grammatically correct and entirely useless. The target readers have no ‘usual’ chain. The phrase describes a system they have never encountered.

Mediating that text means rewriting the reference, explaining the process, and flagging to the client that the original assumed a level of institutional knowledge the new audience does not possess.

That is the kind of work AI cannot do, not because it lacks vocabulary but because it lacks a model of what the reader does not know. Intercultural mediation is a theory-of-mind activity. It requires holding two audiences in mind at once and noticing the gap between them.

The EMT framework places this competence at the centre of professional practice, and CIOL’s own Linguists’ Voices collection makes the same point: “translating true meaning” and “interpretation of nuance and cross-cultural navigation” become more important, not less, as AI tools become more ubiquitous.

 

Transcreation: judgement about what to change, not just how to say it


Transcreation is often described as creative translation, which makes it sound like a luxury service for advertising agencies. In practice, it is something more ordinary and demanding: the exercise of judgement about which elements of a text are ‘load-bearing’ and which are ‘decorative’.

A machine can produce a target text that is faithful to the source at the level of clause and sentence. What it cannot do is decide that a particular metaphor will alienate the target audience, that a cultural reference will read as obscure rather than clever, or that a regulatory phrase must be preserved verbatim even though it sounds awkward. Those decisions require a view of the text as a whole and a theory of what the client is trying to achieve.

A useful rule of thumb: before translating anything, establish which parts of the text are fixed and which are negotiable. Are certain phrases legally required? Are they brand language? Or are they simply how the previous author happened to phrase something? Transcreation is the work of operating within those constraints while still producing something that works in the target culture. The EMT framework’s emphasis on “strategic adaptation decisions” captures this well: the human judgement lies not in generating alternatives, but in knowing which alternatives are acceptable.

 

Data curation: the competence nobody warned me about


Data curation is the competence that would not have appeared on many translators’ lists five years ago, and the one that is now used every day. It is unglamorous, involving building and maintaining usable glossaries, term bases and translation memories. It means separating authoritative client terminology from legacy noise. It also means noticing when a machine translation tool has learned a term from a contaminated source and producing a correction that persists beyond the current project.

The EMT framework includes data curation among the career opportunities that a modern translation education should open up.  Practically, it is the ‘unobtrusive’ infrastructure of quality. When a client sends a 40,000-word technical file and asks for consistency across five languages, the value a professional translator adds is not in typing translations faster than a machine. It is in building the terminological architecture that makes consistency possible, and in maintaining the judgement to know when a term should be standardised and when context demands variation.

This is also where the ethical dimension of AI use becomes concrete. CIOL’s work on AI in public services found ‘widespread but risky use’ of machine translation tools in healthcare, police departments and legal contexts, often without institutional awareness or policy frameworks. Data curation is one of the practical answers to that risk. A curated, verified term base is a form of quality assurance that ‘travels’ with the text. ‘How a professional translator turns a one-off correction into a systemic safeguard’ expresses the mechanism of this base.

 

Quality judgement: the competence that holds the others together


If intercultural mediation, transcreation and data curation are the specific skills, quality judgement is the disposition that connects them. It is the ability to look at an output, whether produced by a machine or by a human, and decide whether it is fit for purpose in this context, for this reader, with these consequences.

CIOL’s chief executive Sara Robertson put this more directly than I can in Linguists’ Voices:

“Professional judgement is worth dwelling on, because it’s precisely what AI cannot replicate. A machine can produce output that looks fluent and indeed plausible. What it cannot do is weigh up competing considerations, recognise when something feels wrong, understand the consequences of a mistranslation in a particular context, or take responsibility for a decision”.

That paragraph describes a translator’s working week better than any competence grid. The machine produces a plausible sentence. The translator reads it and feels a sense of alarm: the register is wrong for this audience, or the term is technically correct but odd, or the sentence is grammatically smooth but implies a commitment the client did not intend. The alarm is not mystical. It is the product of training, experience and a learned habit of reading for consequences as well as for correctness.

The EMT framework’s ‘stronger focus on critical thinking and ethics’ is not a supplement to the technical content. Rather, it is the organising principle. Quality judgement is what turns a set of discrete competences into professionalism. It is also what clients are buying when they commission a human translator rather than running a file through a tool.

 

What has become redundant, and what a new translator should focus on


The skills learned at the start of a career have not degraded, but some have changed status. The ability to produce a fluent first draft quickly is less distinctive than it was. The ability to spot a terminological inconsistency across a 30,000-word document remains very useful, but it is increasingly mediated by tools that can flag repetition.

What has become redundant is the idea that translation is primarily a text-production activity. The framework’s shift from ‘translation’ to ‘transfer competence’ is a signal in the same direction: the professional is not the one who generates the target text, but the one who takes responsibility for the transfer of meaning across a cultural and communicative gap.

A new translator today could usefully focus on three things.

  1. First, learn to read for consequence. Ask what happens if this sentence is misunderstood, and let that question shape your attention.
  2. Second, build curation habits early.  A glossary maintained properly is worth more than a dozen flawless translations that cannot be reproduced.
  3. Third, treat intercultural mediation as a craft, not a personality trait. It is a set of observable decisions about what to explain, what to adapt and what to exclude.

 

Professional accountability and judgement: the ground upon which we stand


The EMT framework is a document for universities, but its implications go far beyond traditional academic dimensions. What it demonstrates to the profession is clear: the skills AI makes more valuable are the ones that were always at the core of good translation: the ability to notice cultural nuance, to judge and to take responsibility.

Tools have changed the ‘mechanics’ and ‘economics’ of text production, but they have not changed the nature of professional judgement and accountability. That is not a consolation prize, but the ground on which the profession can stand.

 

 

Tamer Osman, MCIL CL is a Senior Lecturer in the Department of English Studies in the School of Foreign Languages at Shandong University of Technology, China. He is on LinkedIn at https://www.linkedin.com/in/tamer-osman-38736013a/ and on the web at https://languageslavie.blogspot.com/

 

 

 

Views expressed on CIOL Voices are those of the writer and may not represent those of the wider membership or CIOL.

 

 

 

 

 

Selected References