This post is based on a paper I recently wrote. I was responding to a request from the Welsh Commissioner to imagine alternative possibilities ahead of the Association of Translators conference in Aberystwyth on June 23, 2025.
Summary and disclaimer
I would like to emphasize that I am not a translator nor do I wish to speak on their behalf. I have worked closely with translators for a long time and I foresee:
- content roles (including translation) changing in the near future
- technology changing the focus of content and translation roles – and that people in those roles need to start defining their roles before anyone else does it for them
- that there are opportunities to use the expertise and capabilites of language specialists to improve service content in Wales, by enhancing content in both languages
- that, in addition to translating text and interpreting, there is a chance to cultivate a third pillar for the work of translators, namely bilingual design or multidisciplinary translation (I don’t know what to call it!)
Context and challenge
Machine translation (MT) from English to Welsh has improved significantly, and while this is a technical achievement, it also presents practical, cultural and linguistic risks. If left unchecked – and if we don’t adapt our approach – AI could reduce the Welsh language to a mere translation layer.
In the context of designing user-centred public services, this risks ignoring the worldviews, identities, and lived experiences of the people who use those services in Wales.
Risks of AI in bilingual public services
Unchecked, AI could undermine the Commissioner’s vision of a bilingual Wales where people can live their lives fully in Welsh. It could also undermine the aims of the Digital Service Standard for Wales:
- understanding users and their needs
- everyone can use the service
- service should be designed in English and Welsh.
It could lead to:
- a shortcut mentality: AI is often seen as a shortcut to compliance in all areas, bypassing the deeper work of bilingual service design and understanding user needs. Services will be designed in English and regurgitated into Welsh.
- embedding formal language in all areas: AI could introduce and accelerate technical jargon and formal language that alienates everyday users in both Welsh and English – because this is what it learns from limited existing datasets and corpora. This is a bigger risk than ‘incorrect Welsh’; there is evidence that usability more important to users than correctness and formality
- loss of usability: The issue is not just accuracy, but usability—how Welsh is used in real life, by real people, in real services.
Opportunities for designing bilingual services
In CDPS we believe that translation is not just about communication and compliance.
It’s also about how we deliver effective bilingual services by understanding the needs of all people of Wales and designing services that work for users in both languages. This is not just ‘translation’.
This means:
- understanding the needs of different service users
- considering how services are designed to meet those needs
- continuously improving services in both languages
- designing services bilingually from the start
The translation profession will need to discuss and decide its own future. We are not suggesting or directing how things should change. We’re merely suggesting possibilities for various professions to proactively evolve in the age of AI. In this context, translation as a profession may not be able to exist merely on the plane of the written word, as AI will become more and more capable of taking the load of that work – for better or worse. The important thing is to be proactive and prepared for any changes.
Here are some ideas on how the role could evolve around the needs of public services in Wales and support the Welsh Language Commissioner’s vision of a bilingual Wales where people can live their lives fully in Welsh.
Translators as strategists and designers, who can help service teams create the right content to meet users’ needs in both languages. They could help service designers think strategically about what it means to design a service bilingually.
Translators as editors of AI and machine translation. They could carry on working with the assistance of AI and machine translation on formal, large-volume translation tasks which might not be as relevant to the population at large (that is, they’re not services people interact on a day-to-day basis, such as:
- Policies and policy guidance
- Internal documents
- meeting minutes
- reports )
As technology evolves, this work might become quicker, releasing time for other activities.
Translators as user-centred experts. They could:
- consolidate their positions as the guardians of – not just linguistic standards – but linguistic usability and clear language
- learn, as content and service designers do, directly from user experiences
- be advocates for the language that the people of Wales use and understand when they use services
- help build user-centred terminology, aligning other teams around best practice
- working in a multidisciplinary way by collaborating with user researchers and service designers to make sure that services meet the needs of Welsh language users
These possibilities are similar to other shifts that are being discussed in other content, design and communication roles. We believe translation must be part of this conversation too.
Compliance and usability: considerations for the Commissioner
We also believe these ideas would need careful consideration and guidance from the Welsh Language Commissioner to balance the needs of meeting compliance and of creating a bilingual Wales where people can live their lives fully in Welsh. These 2 things may not always be mutually exclusive.
Service teams across Welsh and UK institutions have repeatedly mentioned their struggle to balance the sheer volume of demand on translators’ time with getting resources to support the bilingual design of services.
The public is more likely to interact with public-facing services than corporate, official or internal documents. People often don’t have a choice but to use some services – such as grant applications, planning applications, council tax reduction etc.)
This raises complex questions about:
- the value we attach to the translation profession and their unique skills to support the Commissioner’s vision. How might technology support this in an ethical and considerate way?
- reimagining what being fully compliant and providing usable services with the Welsh Language Standards will mean in practice
- acceptable ways to fully comply with Standards whilst also prioritising public-facing digital services that support the vision a bilingual Wales where people can live their lives fully in Welsh. That is: would translating internal, corporate docs primarily with technology be an acceptable trade-off if that meant that public-facing services were user-centred and easy to use in Welsh?
Prioritising accessibility
A key priority should be to use AI capabilities to make accessibility tools work better in Welsh.
What ties all of this together is the simple idea that everyone in Wales, no matter their abilities or background, should have equal access to services and opportunities in Welsh.
How might we:
- improve text-to-speech tools so that they work seamlessly in Welsh
- improve speech-to-text tools to:
- automatically create video captions in Welsh and English
- automatically create Welsh alt text for images
- transcribe broadcasts and archives