The Novimed Journal · Clinical research
The Future of Linguistic Validation: AI Workflows and Human Oversight

In his International Translation Day 2026 presentation, Dr. Doraid Matani describes a future in which AI coordinates bounded linguistic-validation tasks while specialists retain control of meaning, cultural validity and final approval. The recording includes a conversational ValiTracks demonstration and a discussion of how professional responsibilities may change.
Dr. Doraid Matani · Published 2026-10-01 · 49:15. Watch with captions and transcript.
Explore the presentation chapters
- 0:00 — Opening and introduction
- 1:01 — Dr. Doraid Matani and ValiTracks
- 3:02 — The linguistic-validation workflow
- 3:35 — What AI can and cannot do
- 4:46 — Translation and reconciliation
- 5:54 — Conversational workflows
- 9:35 — Subject-matter expertise
- 10:57 — Future AI-assisted workflow
- 13:15 — Translation and quality assessment
- 14:55 — Continuous QA and eCOA screen review
- 17:27 — Cognitive debriefing
- 18:22 — Professional roles
- 19:34 — Accountability and human control
- 21:07 — ValiTracks demonstration
- 29:37 — The translator’s role
- 31:06 — Predictions
- 36:34 — Audience questions
About the presentation and speaker
The presentation was delivered for ProZ.com International Translation Day 2026 and published on NovimedGroup LLC’s YouTube channel on 1 October 2026. The recording runs for 49 minutes and 15 seconds, including audience questions. Dr. Matani introduces himself as a medical doctor, linguist and founder of NovimedGroup and ValiTracks. The company’s existing biography also identifies his medical and research background.
This page is an edited thematic account of the presentation, not a verbatim transcript or a regulatory guideline. The recording contains opinions, predictions and a trial software demonstration. Those should be distinguished from published methodological recommendations and from a commitment that every demonstrated or proposed feature is currently available to every client.
A workflow argument rather than a translation contest
Early in the talk, Dr. Matani shifts attention from choosing the best translation model to understanding how AI interacts with the successive activities of linguistic validation. His examples include translation, reconciliation support, quality assessment, electronic review and cognitive-debriefing documentation. The recurring question is where technology can reduce administrative work while preserving the decisions that require linguistic and clinical expertise.
He places a clear boundary around the purpose of the work: assessments will be used by people, so their understanding and human approval remain essential. This is a useful way to evaluate a proposed workflow. Faster generation is relevant only if the resulting language and evidence remain suitable for the intended population and assessment.
Translation support does not transfer decision authority
In the discussion of forward translation and reconciliation, Dr. Matani describes AI drafting and comparison as tools for the linguist. He distinguishes suggestions about competing versions from responsibility for producing the reconciled wording. He also discusses back-translation drafting while retaining linguistic interpretation of what the result means.
The distinction reappears in the audience questions. The system may identify potential issues and provide a wider picture for reviewers, but it does not take over the reconciler’s judgment. Teams considering this approach should therefore ask how suggestions are presented, how a reviewer can reject them and how the final rationale is retained. These are practical implementation questions arising from the talk, rather than claims that one specific method is sufficient for all instruments.
A practical reading lens
Define the concept
Review the evidence
Record the decision
Conversational access with controlled context
The presentation describes a move toward interacting with specialized software through a conversation. Dr. Matani argues that linguists should be able to focus more on the task and less on learning menus. He places that interface alongside controlled resources such as concept documentation, terminology and translation memories, with permissions, timestamps and versioning still present.
The demonstration begins around 21:07. It shows a trial workflow involving language selection, source resources, an AI-assisted draft, a subsequent human version and different access for reviewers or clients. The example is presented as a preview. It should not be read as evidence of a client relationship, production deployment or a universal feature-availability date.
Continuous QA and eCOA review
Around 14:55, the talk extends the same architecture to quality assessment and electronic screens. Dr. Matani describes comparing displayed content with approved bilingual wording and linking potential issues to implementation files. He also explicitly rejects the idea that AI approval alone is enough: a person still needs to walk through the result.
The operational implication is that issue detection and issue resolution should remain separate. A tool can help locate a discrepancy and connect it to the correct object; a qualified reviewer assesses its meaning and the proposed correction. The final implemented version should still be checked. This interpretation is consistent with the talk’s emphasis on making coordination faster while retaining human control.
Cognitive debriefing remains evidence interpretation
The presentation discusses transcription and drafting of interview summary reports as possible forms of assistance. Dr. Matani places the interviewer’s value in understanding the participant’s difficulty, interpreting the evidence and deciding what follow-up questions or suggestions are needed. Preparing paperwork is different from understanding what an answer reveals about the item.
In the question-and-answer discussion, he also describes exploratory work involving parts of interviews while retaining an interviewer’s management. This is described as testing, not an established replacement for human-led cognitive debriefing. The page therefore does not present autonomous interviewing as a recommended standard or an available service promise.
How professional roles may evolve
At approximately 18:22, Dr. Matani outlines changing responsibilities: translators remain language decision-makers, reviewers focus on risk, project managers spend more attention on the process, and interviewers interpret evidence. He anticipates more immediate client involvement through a connected workflow. These are predictions about how work may be organized, not evidence that the underlying responsibilities disappear.
Later, he discusses how linguists should be recognized for identifying difficult conceptual or cultural errors. His comments on remuneration are an invitation to professional discussion, not an announcement of a specific NovimedGroup pricing or compensation model. The central theme is that expertise remains valuable even when generating a first draft becomes easier.
Accountability and the ability to stop
The talk repeatedly returns to accountability for conceptual decisions, approved wording, versions and patient-facing risk. Dr. Matani cautions that a system can make its output look more settled than the evidence warrants. He describes the need for people to intervene, stop a process or return to a traditional approach, including situations involving rare languages or diseases.
For a team designing an AI-assisted workflow, this raises a concrete governance question: can the responsible reviewer actually change the course of the work? Human oversight should include access to the evidence and authority to act on it. A required approval click has limited value if the reviewer cannot see uncertainty or reopen an earlier decision.
What to take into a project discussion
Use the presentation to frame questions about your own programme. Which activities are suitable for assistance? Which source materials are authoritative? Who reviews each output? How are permissions and versions controlled? What happens when evidence is incomplete or the model is wrong? Ask for the current configuration and available capabilities rather than assuming a preview applies unchanged to a live project.
NovimedGroup can discuss a COA/eCOA project in terms of its instrument, languages, population, implementation and review needs. ValiTracks is introduced here as the technological context for Dr. Matani’s perspective. The commercial service remains linguistic validation with an agreed scope, qualified specialists and documented responsibility for the final deliverables.
Sources and editorial scope
Practical workflow recommendations are editorial interpretation. The cited publications retain their own scope and evidence limitations.
Speaker profile: Dr. Doraid Matani
A worksheet for your next project
Capture the instrument version, language-country pairs, review responsibilities and delivery evidence before agreeing the scope.
Download the project brief (plain text)Put the discussion into practice
Discuss your COA/eCOA project
Explore our linguistic validation services and agree the languages, review stages and documentation your project needs.
Request project support →