07/16/2026  |  Hilde Vanaken  |  Applied Clinical Trials

https://www.appliedclinicaltrialsonline.com/view/making-communications-resonate-wherever-people-own-words  

The traditional global-to-local communication model creates fragmented content through sequential adaptation, but a local-first approach using structured data and AI-orchestrated generation can embed stakeholder terminology and preferences from the start, reducing rework while improving clarity and relevance.

Communication is often underestimated. It is not just about sharing information—it is about ensuring that what is intended is actually understood. When that alignment breaks, people don’t just miss details; they fill gaps with assumptions. In regulated environments, those assumptions can translate directly into operational inefficiencies, quality risks, and potential impacts on patient safety.

In clinical research, the importance of clear communication is amplified. Trials bring together a diverse network of stakeholders—participants, investigators, site staff, ethics committees, health authorities, and sponsor teams—each with different expertise, expectations, and perspectives. Information must simultaneously meet scientific rigor and be understandable in plain language, often across multiple cultures and languages. Ensuring that meaning is preserved—rather than lost—across that complexity is far from trivial.

This challenge is intensifying. Clinical trials are becoming more complex. They incorporate digital tools and artificial intelligence into already intricate workflows—introducing concepts that can be difficult to fully grasp, even for highly trained professionals.1 This raises a critical question: if even experts need time to interpret and understand these advances, what does this mean for participants, physicians, and other local stakeholders—who rely on clear, locally relevant communication to act and decide with confidence?

Equally important, the industry faces a growing paradox: as organizations intensify efforts to standardize content globally, how can they preserve local communication styles, language nuances, and cultural context that are essential for true understanding?

It’s time to pause—and fundamentally rethink the entire model for developing stakeholder-specific communication in local languages.

In this article, we explore the current global-to-local model for developing stakeholder-specific materials, its key inefficiencies and challenges, and how a fundamentally different, local-first approach can redesign communication at its foundation.

The hidden complexity of stakeholders’ unique local needs
Across the life sciences industry, the development of stakeholder-specific materials in local language—whether for clinical trials or commercial use—typically follows a linear “global-to-local” model. A global document is authored centrally, usually in English, then translated and adapted into country-level versions, and finally further tailored at site level.

In theory, this approach creates consistency. In reality, it introduces fragmentation.

Each step in the process involves different contributors across global, country, and site levels—each one bringing necessary expertise, but also a specific lens. As content moves downstream, it is repeatedly interpreted, adapted, and reshaped—often without full visibility across the end-to-end process.

The core challenge is simple but profound: there is no one-size-fits-all. Every company, country, and site operates within its own set of requirements, cultural context, and communication style. What is considered clear and acceptable in one setting may be unclear or even unacceptable in another. Even seemingly simple elements—terminology, phrasing, or tone—can vary significantly across stakeholders.

As a result, the industry faces a persistent challenge. Organizations strive to standardize content globally to improve consistency, quality, and efficiency, while simultaneously needing to adapt communication to local requirements, communication styles, language nuances, and stakeholder expectations. The current model attempts to bridge this gap through repeated rework—fixing and correcting content at each stage rather than designing it right from the start.

Consent as a lens into systemic complexity
Informed consent provides a clear illustration of this challenge (Figure 1). It is a foundational process in clinical trials—without informed consent, there are no participants, and without participants, no trials. It must communicate critical elements such as purpose, procedures, risks, benefits, and data use in a way that participants can understand and act upon.2,3,4

Consent must meet regulatory and ethics committee requirements, while also reflecting local context, site practices, and local language nuances. Differences between global, country, and site materials of the same study can be significant:

• added, modified, or removed titles, sections, phrases and wordings
• different document layouts, formatting, structure and data collection needs
• stakeholder-specific preferred local language terminologies and writing styles

These local nuances can make the difference whether participants truly understand the information—the ultimate goal of consent. However, insight into these local specifics and variations between global, country or site versions remains limited. Content is rarely reused across studies, and feedback from sites is not systematically captured—resulting in repeated rework and growing frustration at the site level.

Figure 1. The Fragmented Global-to-Local Consent Development Process

Source: Hilde Vanaken

The operational footprint of creating consent documents is substantial. For a single Phase III study, the process can involve thousands of documents, input from more than 60 different departmental subject matter experts, timelines of several months, and costs reaching seven figures. This same process is repeated for each study, often involving different contributors.

At the same time, digital formats and multimedia (~eConsent)—an emerging trend to improve consent understanding—further amplifies this already inefficient development process, adding more resources, increasing costs, and extending timelines.

The flexibility gap
Beyond these inefficiencies, there is also a usability challenge. Today’s processes often produce documents that are rigid—they do not easily support different ways of engaging stakeholders.

For example, sites may want to guide participants through information step by step—starting with a concise summary before presenting full details. While this aligns with improving understanding, current document models rarely enable such flexibility without additional manual effort.

Consent is just one example, but it reveals a broader truth: stakeholder-specific materials should be driven from the local perspective, at the level of sites and participants where understanding is created, while still preserving the benefits of sponsor standardization.

Re-inventing the global-to-local process at the foundation
Addressing these challenges requires more than incremental improvements. It requires rethinking the model itself. Rather than starting with a global document and adapting it downstream, the process should be inverted: local stakeholder-specific needs—how they communicate, the terminology they use, the formats they prefer—should be embedded from the start.

A two-step model: Build & generate
This structured foundation—referred to as a Local Data Engine—can be summarized in two stages:

• Build stakeholder-specific local and global intelligence

A one-time activity in which a limited set of historical study documents is automatically extracted, categorized, back translated, and mapped between all stakeholders into a reusable knowledge base—captured as structured data and blueprints in both local language and English.

• Generate & update stakeholder-specific study documents

New stakeholder-specific study documents are dynamically generated and updated by combining study-specific data—derived from global study inputs—with the local intelligence of the targeted countries, sites, and languages of the new study.

The Local Data Engine extracts metadata, titles, text, layouts, as well as stakeholder-specific terminology and descriptions related to study designs (e.g., randomized, open label), therapeutic area terms (e.g., compound description), and other elements, both in local language and English.

It also maps how content and structure are linked across various documents: country to global, and site to country. Translation and back translation incorporate preferred stakeholder terminology in local language. For example, back-translation to English will be aligned to sponsor-preferred global English terminology (e.g., consistently using “study physician” where that is the agreed global term).

Blueprints are also not fixed in stone. They can be continuously refined, bringing in more personalization and improvements in line with local stakeholder needs.

Artificial intelligence (AI) agents orchestrate this process—automating extraction, categorizing, translation, mapping, and generating of stakeholder-specific blueprints and study documents in local language, while ensuring consistency and traceability across global and local levels.

Figure 2 illustrates the Local Consent Engine as an example of a Local Data Engine.

Figure 2. The Local Consent Engine: a Two-Step Model to Build Local Intelligence and to Generate Study-Specific Informed Consent Documents (ICDs) in Local Language (Illustrative Example of a Local Data Engine).

Source: Hilde Vanaken

What changes in practice
Embedding local intelligence at the foundation unlocks several capabilities:

• Parallelization instead of sequencing: Local requirements in local language are addressed from day 1, rather than through successive handoffs.
• Reuse instead of rework: Preferred language, phrasing, and structures are captured once and applied consistently.
• Visibility instead of opacity: Sponsors gain transparency into how content varies across contexts - enabling consistent deployment and improvement at all levels.
• Flexibility by design: Content can be tailored dynamically to different needs—for example, generating both summary-level and detailed versions without duplicating effort.
• Efficiency gains at scale: Significant impact on cost, timelines, resources, and quality—combined with an improved stakeholder experience through increased content flexibility

Once content is structured as data, expansion into multimedia and digital formats, or data collection becomes a natural next step. This will reduce many of the current challenges associated with multimedia and data collection and enable greater flexibility for end users.5

Transforming the broader clinical research ecosystem
This model not only reshapes sponsor processes but can also deliver value across the clinical research landscape. Stakeholders such as ethics committees and health authorities can rely on structured, reusable content and requirements, reducing the need to repeatedly review similar content from the same sponsor across studies, while improving both efficiency and consistency. Requirements from these bodies can also be embedded directly into structured blueprints.

Patient and site organizations can benefit in a similarly meaningful way. Today, their perspectives on patient communication are typically captured in an ad hoc, fragmented, and study-specific manner. With this model, their input can be centrally captured, structured, and reused consistently, ensuring that local context and patient-centric considerations are more systematically reflected across studies.

In essence, the model shifts from a sequential global-to-local document-driven approach to a parallel, local intelligence-driven model. Stakeholder-specific requirements, preferences, and communication nuances are embedded from the start—while still preserving the benefits of sponsor standardization.

Shifting to a novel end-to-end operational model
The Local Data Engine also enables a novel operational setup—shifting from a fragmented, multi-department process with multiple SMEs to an integrated, end-to-end model with a single role overseeing stakeholder-specific material development in local language for a defined process.

This end-to-end role combines:

• process-specific domain expertise, bringing together capabilities traditionally distributed across multiple departments
• data and AI capabilities, including data management, structured data handling and effective interaction with AI-driven systems
• orchestration of AI agents, leveraging automation, confidence scoring, and continuous learning to drive consistency and quality
• targeted collaboration with departmental SMEs, engaged on an ad hoc basis for specialized input where needed

Crucially, this is not about eliminating human expertise—it is about concentrating it. It enables experts to focus on what matters most: judgment, context, and engagement.

Figure 3 illustrates the impact using consent as an example. Today’s consent process spans multiple sponsor departments and experts across global, country, and site levels—clinical operations, medical writing, physicians, privacy, legal, human biological samples, patient and site engagement, and others. In the new model, we shift to a single end-to-end role, supported by AI agents, enabling up to 70% reduction in departmental SME efforts.

Figure 3. Next Generation End-to-End Operational Model (Consent as an Example).

Source: Hilde Vanaken

Introducing the NextGen local digital data flows
Traditional document processes treat content as static text—created, reviewed, translated, and managed as separate artifacts.

Industry initiatives such as ICH M11 structured data protocol guideline6 and TransCelerate/CDISC Digital Data Flow7 are already moving beyond this paradigm. They fundamentally change the role of documents from static text into structured data and reusable intelligence—create once, use many times—or from an information-driven to a data-driven ecosystem as noted by the EMA.8

The Local Data Engine takes it one step further: extending from global sponsor documents in English to stakeholder-specific material in local languages. It not only captures stakeholder-specific requirements, writing styles, institutional practices, and preferred local language expressions as structured data elements in both local language and English, but it also maps them consistently across all documents (Figure 4).

Figure 4. Next Generation Local Digital Data Flows (Consent as an Example).

Source: Hilde Vanaken

As a result, global consistency and local relevance are no longer competing forces. They become part of a single, harmonized system, where stakeholder-specific communication is generated from a shared, structured foundation.

Moving beyond translation—embedding stakeholders’ own language style
Translation has traditionally focused on linguistic accuracy—ensuring that words are correctly converted from one language to another. While essential, this is only part of the equation.

True understanding depends on more than correctness. It depends on whether communication reflects how stakeholders naturally interpret and use language within their specific context, practices, and preferences.

In practice, even seemingly equivalent terms can differ in meaning and familiarity across stakeholders. One stakeholder may use “investigational drug,” another “medicinal product,” and another “research drug.” While all translations may be technically correct, they are not necessarily aligned with how stakeholders are used to communicating. Sites often adopt internal terminology that participants are familiar with. Deviating from this—however linguistically accurate—can reduce clarity rather than enhance it.

In today’s environment, large language models (LLMs) are being rapidly adopted, accelerating both content generation and translation. While powerful, these models introduce important risks when used in isolation. Outputs may be linguistically fluent but misaligned with local expectations, terminology, or intent. Nuances can be diluted, and critical context may be lost.

The Local Data Engine addresses this by combining LLM with controlled translation using stakeholder-specific language assets derived from the local intelligence layer. For example, when generating a consent form in Belgian Dutch for a specific site, the system applies the predefined terminology and language preferences captured for that site and language during the local intelligence build phase.

In this model, the objective is not perfect, generic language - it is effective communication: aligned with how each stakeholder expects to receive, interpret, and act on information (Figure 5).

Figure 5. Next Generation Translation Models.

Source: Hilde Vanaken

From concept to scalable reality
While the ideas outlined here might sound transformative, they are not theoretical—they are already being translated into practical, scalable solutions.

A local-first approach ensures that communication is designed around how stakeholders actually think, work, and interact—embedding their needs, preferences, and language from the outset.

Consent was used as an illustration given its critical role in clinical trials, its importance in ethical AI, and the learnings from the cross-industry European Forum GCP eConsent initiative.9 The same model can be used for any stakeholder-specific material—clinical, regulatory, safety, and commercial. What is learned for a specific country or site, such as preferred terminology and communication patterns, can be consistently reused across all materials for that context.

Life sciences organizations vary widely in their processes, governance models, and technology landscapes. Flexibility in the model is therefore recommended, using modular accelerators tailored to organization-specific processes and technology ecosystems rather than applying a one-size-fits-all solution.

The goal is not just to produce better documents. It is to enable communication that truly resonates—wherever people are, and in their own words.

About the author
Hilde Vanaken, PhD, Eng, MsC, is Head of Pharma R&D Transformation, Life Sciences at Tata Consultancy Services (TCS). She has led major transformation initiatives across the life sciences industry, including at Johnson & Johnson, TransCelerate, and the European Forum for Good Clinical Practice (EFGCP), collaborating closely with Health Authorities, Ethics Committees, patient organizations, and site organizations.

References

1. Vanaken H. Rewiring Clinical Research: AI-Augmented Data, Processes, and People. Applied Clinical Trials, 26 June 2025.
2. ICH Harmonized Guideline for Good Clinical Practice E6(R3). Final version, 6 January 2025
3. ICH Harmonized Guideline for Good Clinical Practice E6(R3), Annex 2. Final version, 3 June 2026
4. FDA Informed Consent: Guidance for IRBs, Clinical Investigators, and Sponsors. August 2023.
5. ICH Harmonised Guideline: Clinical Electronic Structured Harmonised Protocol (CeSHarP) M11 Template. Final version, 19 November 2025.
6. CDISC Digital Data Flows (DDF) for Clinical Trial Protocols
7. Arlett P., et al. Better data for better medicines: a path to data-driven medicines regulation. Nature Reviews Drug Discovery, 29 May 2026.
8. Vanaken H., et al. Effective eConsent Strategies for Every Study: Utilizing the eConsent Fit-for-Purpose Study Framework. Applied Clinical Trials, 12 August 2024.
9. European Forum for Good Clinical Practice (EFGCP) eConsent initiative