RWS’ Daniel Brockmann and Luis Lopes opened the 2026 ETUG Bitesize series. They presented the current Trados Roadmap (as of late February 2026). Since their presentation at ETUG 2025, the industry shift has continued unabated. The frenetic pace of advances in AI now influence development cycles as AI agents rather than developers write code. AI has triggered a fundamental change regarding translation: use cases are moving away from established sentence-based segmentation and translation practices.
New generative translation approaches no longer send individual segments to MT that are returned without any context. Instead, it now sends translation memory and terminology data as well as document summary data. Doing so provides greater context at project level and makes use of custom prompting. The same custom prompting approach allows different translations to be delivered for different audiences. For example, different translations of a medical nature could be delivered to doctors and patients respectively.
RWS’ Translation Technology Insights report in 2025 identifies top use cases regarding AI as relating to quality enhancement and automation of translation processes. So it is a case of quality over increased quantity. These developments also focus on greater integration with other systems that companies use. The approach is to shift viewing translation as a cost centre, to a driver of growth. In this regard, Trados now integrates over 11,000 LLMs, as well as an out-of-the box LLM.
Applying corporate style guides to Generative Translation
The generative translation approach also allows corporate style guides to be adhered to. Rules can be applied to models to reflect tone of voice, using a technique known as Parameter Efficient Fine Tuning (PEFT). PEFT adjusts only a small subset of parameters rather than retraining a full model. This has the advantage of requiring reduced computational costs and training time of the underlying LLMs, while also fitting a company’s tone of voice. Other aspects of generative translation involve using retrieval-augmented generation (RAG) processes e.g. as translation memory and terminology matches change. While this was previously only possible on a segment-by-segment basis, it now takes place using larger batches of translation units.
The new Smart Review feature is a workflow task within a project. It reviews and provides feedback about the quality of segments (e.g. punctuation, grammar, incorrect terminology) in the online editor. This was previously available interactively only at active segment level. Now it is part of the workflow of setting up a new project, so that it can initially address the most problematic segments, before looking at the others. Customers are currently testing a routing approach that skip the translation stage and to head straight to the review phase. This is possible if certain thresholds are not breached during the smart review.
Smart Insights is Trados’ localisation or project management copilot. It can be asked questions regarding reporting data, for example to generate reports such as ongoing translation tasks sorted by owners (such as what is translator X working on at the moment, what do they have to deliver in the next week etc.), and rather than a static report-based approach allows a more conversational approach to interacting with the data.
Moving away from the language services silo
The real change is that Trados is now an AI orchestration platform. This allows connectivity throughout a whole company, rather than siloing language data within language services or localisation departments. The aim is to help prevent colleagues from bypassing any bottlenecks by using alternative “shadow translation” solutions. Such solutions like Google Translate or ChatGPT often fail to take terminology and tone of voice into account. This is achieved by translation requests being made to the Trados as a platform rather than to traditional language services.
RWS has built a Trados Agent together with Microsoft that can be integrated into M365 Copilot and Teams. The Trados Agent is available in the M365 Agent store, allowing users to use to the Agent’s interface to talk to the Trados tenant, by inviting the Trados agent into the conversation. It can also be executed by having a chat with yourself in Teams. Currently it can be used for translation and linguistic assessments through Teams. It uses your TMs and terminology rather than the output from a generic LLM. In addition, it removes the shadow translation risk whereby staff members bypass internal language services and localisation services. The upshot is that languages services are brought into the conversation over corporate AI adoption.
Bringing Content Owners and Subject Matter Experts to the party
In layout editing is a new feature of the Online Editor that focuses on involving other users (content owners and subject matter experts). Previously it was only available for Word and PowerPoint, but now also covers HTML and XML including stylesheets. Content owners and subject matter experts can then edit and review in a format that looks more like a WYSIWYG editor. This avoids being exposed to the “segment shock” in the Trados environment. Presenting output in a format that they can interact with speeds up feedback and makes it easier to adopt the changes. Content Owners and Subject Matter Experts’ level of input may involve leaving comments or manually changing a translation inline.
There have also been usability enhancements, to allow more maintenance tasks in the online editor, such as deleting and editing translation units directly from the interface rather than having to go to the TM maintenance pane. This is available on a permissions-based basis and is only open to users with “batch edit” permission, which is tied to TM maintenance. Fuzzy match filtering has been improved – from a binary yes/no to allowing special ranges of fuzzy matches to be displayed.
Improvements in Terminology
There are now multiple levels of terminology subfields possible. Finer access levels also have been added at entry level (e.g. as public, draft, confidential) or language level. This allows users to only edit entries, in certain languages and to use specific different types of user groups. There is also a new AI Terminology Extraction field for starting monolingual/bilingual terminology extractions. Bilingual extraction will be likely to be available from sometime in Q2. It is also intended that this feature will be able to conduct terminology extraction from cloud-based TMs.
AI Terminology Extraction is likely to require an additional subscription tier. There may be a small introductory free tier, and then subscribers purchase additional volumes of AI tokens for this purpose. Terms extracted in this way can be added to a new terminology database or an existing one, as well as defining synonyms. Domains (e.g. legal, financial etc.) can be defined too. Context information, frequency of use, parts of speech can be added by the LLM in Trados cloud. Custom fields were previously at project level, but are now possible at vendor and template level. Bilingual exports from cloud termbases have not yet been implemented, but are scheduled for future implementation.
Security features and reporting:
IP allow-listing has been introduced to prevent unauthorized access to a Trados tenant by users outside of your organisation, which is also useful for mitigating brute force attacks. This is intended for use by people working in-house or via a VPN using static IP addresses. This is currently only available upon request. Trados Data Bridge can be used to pull out data from Trados into business intelligence tools (e.g. Microsoft Power BI). New additions and improvements for custom reporting have been added – including a field about whether a translation has been started.
What is coming up in Q2:
Since the Bitesize session took place, RWS has announced Language Weaver Pro – a new AI translation model built in partnership with Cohere. Managerial staff are exploring AI as a basis of reducing costs and trying to increase profits. There is a greater technology focus, with providers pivoting to AI to remain competitive. A key development has been that customised tools with external partners reach deployment far more frequently than internal agent developments.
There is a new vendor selection view that is scheduled for general release in Q2. This will allow the defining of vendors for translation or linguistic review, also in different language pairs. In addition it can then calculate the respective costs, to allow you to decide which vendor to assign a job to. Currently customer quotes and vendor quotes are separate. It will be possible to include the vendor quotes in the customer quotes in the future.
Concurrent editing: It will be possible for multiple Online Editor users to working on a single large document – permitting parallel peer review while translation is ongoing, or for multiple customer reviewers (e.g. a mechanic or the legal department) to work on a single document and multiple users to work on customer feedback in the online editor. This is particularly useful, where there is a very tight deadline which does not permit sequential translation and review cycles, rather than having to split up documents. Tasks can be assigned to specific users within groups. The limitation is that two users cannot work on the same segment at the same time. This feature has already existed in GroupShare.
Trados “Classic” / Trados Studio / Trados GroupShare / Passolo
The 64 bit reality!
Trados Studio 2026 aka “Quantum” is the next major release version of Trados Studio, although with a lot of momentum towards subscription-based models rather than perpetual licences (and buy upgrades). Perpetual licences will continue to exist in certain sectors as required, but there is a distinct move towards subscription-based offerings (especially for freelancers). Trados Cloud subscriptions may change in the future to include a certain number of Trados Studio seats in the future. Classic (i.e. non-cloud based products) Trados Studio will continue to be upgraded and updated. Future releases are unlikely to be named after years, after this 2026 release.
The key improvements are: 64 bit architecture (to break free of the limitations of the 32 bit architecture). This will allow users to work with very large files and projects (an example was cited of a file with 487,000 translation units that opens in the Alpha version of Trados 2026, which would not in the 32 bit version – using nearly 7 Gb of RAM whereas 32 bit versions would only be able to access around 3 Gb); to handle bigger TMs, termbases and AI-assisted workflows, while ensuring improved stability under heavy load (out of memory errors should become a thing of the past). The modernisation of the terminology offering will also reflect this change – especially for those users who are using local termbases. These improvements have meant an extended beta testing programme.
Terminology
Local Terminology Modernization: the 32 bit .SDLTB format will no longer be supported and will be replaced by a new local termbase format (.TTB = Trados Termbase) – there will be backwards compatibility for the time being. Terminology recognition and checking have also been improved. Terminology is particularly key for AI translation.
MultiTerm Desktop will stay on 32-bit and is now “fully decoupled” from Studio, i.e. supporting local termbases (.sdltb) and GroupShare. For managing local termbases using advanced features it will be necessary to convert .ttb files back to .sdltb. Filters and layouts will be migrated to Studio over time (Studio will get a Termbases tab). GroupShare termbases will continue to be managed in the same way.
Cloud-based terminology workflows will be improved for adding terms as well as the handling of forbidden terms. Some features that were only possible from the browser will also be possible from Studio (e.g. marking multiple cloud projects as complete).
The forthcoming terminology enrichment feature will fill in missing fields in terminology entries, even as far as sourcing suitable images. Selectable options include providing three definitions. There is no fixed expected deadline for this, and there will also be the option not to have it.
GroupShare, Usability and Passolo
GroupShare integration: there is faster search across multiple TMs on GroupShare and new integration with server-based terminology, so that all the results across multiple TMs will be given in parallel rather than sequentially. Users of GroupShare should move to the latest CU version to be able to use the latest version of Studio.
GroupShare 2020 SR2 was recently released. Language Weaver Edge (on premise) has a web-based Admin Center user interface. GroupShare also fully supports Language Weaver’s MTQE (and lock high-quality translation based on thresholds). This works with a privately hosted LLM. The infrastructure is already on .NET 8 but in the future will move to .NET 10. Following the SR2 release there was also the CU14 release with enhancements to project and project details views, improvements in role and permission management, live TU count updates, more support for new language variants. Trados GroupShare will become Trados Edge. This future development will also include on premise/cloud LLM integration and is expected in Q3. There will be possibilities for hybrid on-premise/cloud ways of working – which might permit MultiTerm Online to be placed into End of Life.
Usability enhancements will see the introduction of support for dark and high-contrast themes in beta (i.e. Studio honouring Windows 11 themes) – such as dark sky, dusk and desert themes.
AI Translation with Context is still under early development. It will take document context into account, with the possibility to set the number of segments to send to the AI to obtain a more contextually-appropriate translations. This will work both in batch pre-translating as well as in interactive editing.
Passolo will also be AI-enabled to be able to work with Quantum.
Remarks from audience questions:
Pricing issues for high volume users and data sovereignty
Some questions asked about pricing issues for high volume use cases. While it may be possible that low volume users have enough volume included as part of their subscription package, high volume users would need to purchase additional volume packs, although there is no definitive approach yet.
Regarding the uncertain geopolitical situation, one question asked about considerations or contingencies regarding the current concerns regarding data sovereignty. The AWS sovereign cloud was mentioned as being located in the EU, and to which no-one in the USA, including AWS staff, has access to.
Being part of the conversation…
Following on from some discussions at ETUG 2025, it was remarked that the issue is often one of “the conversation (is) happening elsewhere”: namely that managerial decisions are frequently being taken regarding language services without involving language services. The question was raised about how RWS is making sure that they are heard in a similar context. RWS’ management is ensuring that they are raising awareness of their products with the major technology consultancy firms and that such offerings are not going to be replaced by AI, as these are also the companies that C-suite professionals listen to. The move away from Trados as a “gated community” that is purely the reserve of language services and localization departments towards being a platform that everyone can access within a company is also important in this regard.
An “AI wrecking ball” scenario exists in some companies. There is a push towards everything AI without really understanding all the implications. AI needs good data, but cannot create it itself, but needs to be extracted from the rich seam of linguistic data held by language services, thereby highlighting their importance. Regarding CAT tool use, it was remarked that while the number of active users has been plateauing, there has not been a massive (e.g. 10-20%) slump. Active users’ output is still increasing, thus demonstrating the productivity enhancement that has occurred from using MT/AI.
Did you miss this session?
By registering for the rest of the ETUG Bitesize Sessions 2026, you’ll receive access to the video of this session and the presentation (password-protected access).
The next ETUG Bitesize will take place on Thursday, 11 June, 2026 from 10:00 – 12:00 CEST and will be on Moving to the Cloud: Real-World Migration Stories – with a presentation by Sara Grizzo and Andreas Merz (Munich Re) and Luis Lopes (RWS).