What we're seeing across regulatory medical writing and AI.

Beyond Drafting Speed: What Actually Matters in AI for Medical Writing

As AI adoption accelerates across life sciences, much of the conversation has focused on drafting speed and content generation. But are these really the metrics that matter most in regulated medical writing?

In this thought-provoking discussion, Barry Drees, PhD, Founder Advisor at TriloDocs and former President of the European Medical Writers Association (EMWA), and Louise Martinsson, PhD, Senior Director and Head of Medical Writing, Hypercare & Enablement R&D at GSK, explore the challenges and opportunities of AI in regulated documentation.

Drawing on decades of medical writing and operational leadership experience, they examine why verification, trust, reproducibility, and governance are increasingly becoming the defining factors in successful AI adoption. The discussion explores the distinction between generating content and constructing regulatory evidence, the growing importance of reducing verification burden and QC debt, and how medical writers can help shape the future of AI-enabled authoring.

Rather than asking how quickly AI can draft, this conversation asks a more important question: what does trustworthy AI look like in regulated medical writing?

The conversations have changed. Here's what large pharma is actually saying about AI.

Over the past few months, I've had a lot of conversations with VP-level and senior medical writing and regulatory leaders across large pharma organisations. A year ago, many of those conversations still centred on whether pharmaceutical organisations should invest in AI for medical writing at all. That debate is largely over. Most large pharma teams are already running pilots, contributing to enterprise-wide innovation programmes, and actively evaluating where different technologies fit. The question now is more nuanced — and far more important:

"What kind of AI actually fits the way regulated medical writing teams work?"

The conversation has moved beyond demos and first-draft speed. It is now about operational fit, trust, governance, and whether AI can reduce workload without creating new risk somewhere else in the process.

GenAI is useful. But GenAI alone is not solving the core problem.

There is genuine openness to generative AI across large pharma — formatting support, low-risk drafting tasks, summarisation, selected repetitive work. But there is also growing realism from people who have tested these systems inside regulated environments.

The issue is not that GenAI outputs are unreadable. Often, they are fluent and well-structured. The problem is that someone still has to verify whether they are right. If AI shifts effort later in the process, onto more senior people, or into a higher-risk review stage, that is not efficiency. That is risk reallocated.

"I don't want to QC content. I want to stop writing it."

The goal is not to generate more content faster. The goal is to remove manual writing effort without introducing a new verification burden downstream.

Many AI programmes still misunderstand what medical writers actually do.

One recurring frustration is that some AI initiatives appear to misunderstand the real work of medical writing. Medical writers are not simply converting tables into prose or polishing language. They are applying judgement — deciding what matters, identifying clinically relevant signals, aligning stakeholders, managing review cycles, and deciding what not to over-interpret. That distinction becomes obvious when tools move from controlled demos into real workflows.

This is why "time to first draft" is becoming a less useful metric on its own. A first draft that requires extensive checking or reconciliation has not necessarily accelerated the process. The better measure is time to a trusted, reviewable draft — one the writer can move forward with confidence.

Table-to-text is still broken — and everyone knows it.

Generating text is relatively easy. Generating text that accurately reflects the data, applies consistent logic across every table and parameter, and holds up under regulatory scrutiny is a different problem entirely. Teams are increasingly distinguishing between three categories of automation:

• Formatting and structure — relatively well handled by existing tools
• Narrative flexibility — useful for drafting support, with appropriate controls
• Data-to-text generation in regulated documents — a distinct and largely unresolved category, judged by a fundamentally different standard

That third category is where the real operational gap sits. The market is moving from "Can AI write?" to "Can AI reduce the work required to trust what was written?" That is a very different standard.

Deterministic output has become non-negotiable.

One nuance I was not fully anticipating is how sensitive teams have become to output instability. Many large pharma organisations use early prototypes and dummy data to lock down messaging and align reviewers well before final results are available. In that workflow, a system that produces different text every time it is refreshed is not a feature — it is a source of friction. Every rerun that produces new wording creates new review work.

Deterministic approaches — where the same input reliably produces the same output — are seen as better aligned with how these teams actually operate. They allow teams to lock down content earlier, regenerate predictably when data updates, isolate genuine changes from noise, and reduce unnecessary re-review. That is not a technical preference. It is a workflow requirement.

Verbose outputs are part of the problem.

Many teams are dissatisfied with the shape of AI-generated content — not just its accuracy. Too often, documents describe small numerical differences at length and only later tell the reader that none of those differences were clinically meaningful. Experienced reviewers want the interpretive point early. Lean writing is not about removing useful information. It is about putting judgement before description.

New tools have to fit into workflows that already exist.

Large pharma teams are not evaluating tools in a vacuum. Many have already built lean authoring frameworks, strategic review models, acceleration pathways, prototype-led workflows, and enterprise AI governance structures. New technology has to fit into that operating model — it cannot simply generate a draft and assume the rest of the workflow will adapt around it. The question is no longer "Can it produce text?" It is "Does it reduce steps, create stable outputs, support stakeholder alignment, and fit how we already work?"

The workflow model itself may be starting to shift.

Right now, many large pharma teams invest heavily upfront in shells, mock data, and prototypes to align stakeholders before real results are available. That process has real value — it helps teams think ahead, align on likely messaging, and reduce late-stage disagreement. But as generation speed improves, some teams are beginning to ask whether the same level of upfront drafting effort will always be necessary.

If a complete, high-quality draft can be produced rapidly once real data lands, the logic begins to shift from prepare → align → wait → rewrite, to generate → review → refine. The more interesting question is whether rapid, deterministic generation from real data can preserve the alignment benefits of prototyping while reducing the amount of speculative pre-data drafting required. Early days — but worth watching.

Adoption is organisational, not just technical.

Even where teams are genuinely interested, the path from pilot to production is rarely straightforward. The questions I hear most are not about capability — they are about fit. What happens after the pilot? What does production look like at scale? How does this sit alongside existing enterprise programmes? How do we make it budgetable, governable, and explainable internally? In large organisations, a tool does not get adopted simply because it works. The conversation has moved from demos to operational viability.

So what does a viable solution actually need to do?

It is not enough to generate text. The systems gaining real traction are those that construct outputs directly from structured source data, apply consistent predefined logic across all tables and parameters, regenerate identically when inputs change, eliminate the need to QC numerical outputs, separate high-risk data processing from lower-risk narrative support, and fit into existing workflows without introducing instability. Less drafting tool. More regulated infrastructure for evidence construction.

Final thought.

The goal is not faster drafting for its own sake. It is faster progression from data to trusted review. That means stable outputs, a reduced QC burden, better table-to-text handling, a clear division of labour between AI and human expertise — and a model that actually fits how large pharma operates.

The goal is not speed. It is speed without QC debt — and without compromising trust in the output.

If this reflects what your team is experiencing, I'd be glad to compare notes. This is a conversation worth having properly.

Stop Using AI. Start Implementing It.

Everyone is investing in AI. Far fewer are implementing it well.

We spoke with Gaspar Wong, Head of Product at TriloDocs, to understand what separates the two and what leaders in regulated industries need to know before they start.

1. Everyone is talking about adopting AI — but there's a difference between using AI and implementing it. Where do most organisations go wrong?

The distinction matters more than people realise. Using AI — giving your team a ChatGPT licence, letting people experiment — is relatively low risk and low commitment. Implementation is a fundamentally different undertaking. It touches your processes, your governance, your people. And that's where most organisations underestimate what they're signing up for.

The most common mistake is starting with the solution rather than the problem. Companies scan the market, find something that looks promising, and roll it out — without having clearly defined what they're trying to fix or how the tool fits into the way their teams actually work. The result is predictable: friction, resistance, and an AI investment that never really lands.

What separates successful implementations is the willingness to slow down at the start. To ask: where are we losing time? Where does repeatable logic eat into work that should require real expertise? Where do we still need human judgment, and where can we safely remove it from the equation? That diagnostic phase is what makes everything that follows coherent. Once you understand the problem, the right architecture becomes obvious. Without it, you're just adding complexity and calling it transformation.

2. There are fundamentally different types of AI — rule-based systems, large language models, agentic workflows. How do you avoid picking the wrong type of AI for the wrong problem?

There are two sides to this: understanding the tool and understanding the problem. And honestly, the first part is the easier one. There's plenty of material out there on what different AI systems can do. What you can't read up on is what to use for your specific situation.

For me, the key point is to tackle the problem at hand with the correct tools and approaches — and the easiest way to think about it is in layers.

At the base, you have tasks that require absolute consistency — structured data processing, compliance checks, calculations where the same input must always produce the same output. Rule-based systems were built for this. Above that, you have tasks that involve ambiguity: drafting narrative content, translating complex findings into lay language, working with context that can't be reduced to a formula. That's where large language models add real value. And at the top, agentic workflows that orchestrate both — powerful, but the most complex to set up and the easiest to misapply.

Where things go wrong is when there is a mismatch. If you ask a large language model to infer from messy structured data, or to make clinical interpretations that should be governed by external logic — ethical considerations in a paediatric trial, risk factors not previously documented — you're asking it to guess where you need certainty. The opposite is equally limiting. The safest approach is to separate your problems first, then build the architecture around that separation. Deterministic systems where precision is mandatory. Generative systems where flexibility is needed. And clear orchestration between them.

3. What does it cost to run AI in production — not the licence fee, but the real operational overhead that nobody talks about upfront?

The licence fee is the easy part. It's a number on a yearly invoice, and everyone understands it. What's harder to quantify — and what vendors rarely walk you through — are the layers underneath it.

The first is integration. How does this system connect to what you already have? How clean is your data, and what does it cost to get it there? Do you have the infrastructure, the technical architecture and backbone system, the IT resources and knowledge to adopt? If not, you'll have to build all of that before you create value with AI adoption. These are real costs, and they're almost always underestimated.

The second is ongoing operation. This isn't a solution you deploy and forget. You're monitoring outputs, running quality checks, maintaining governance and audit trails, managing a human review process. And the underlying models evolve, regulations change, data changes, the technology moves. Keeping pace with that is its own workstream.

The third is exit. It sounds premature to think about it at the start, but it's exactly when you should. What are the vendor's data security commitments? What happens if the relationship ends — by choice or otherwise? How do you disengage without exposure? In regulated environments especially, these aren't hypothetical questions. They need answers before you sign anything.

4. "AI saves time" is the headline everyone leads with. But what does timesaving mean when you factor in integration, change management, and workflow redesign?

It's real — but it's not immediate, and it's not automatic. Many AI tools feel fast upfront but shift the burden downstream. If experts still need to verify every output, instead of saving time, you've just redistributed the risk. The early stages of any serious AI implementation will often look like the opposite of efficiency. Teams are learning, processes are being redesigned, and the tool hasn't yet found its place in the workflow. That's normal, and organisations that expect instant returns tend to abandon good implementations too early.

The gains come once the system is properly embedded. And they're most visible in the work that nobody wants to talk about: the repetitive, high-effort tasks that consume expert time without requiring expert human contextual judgment. Validating data points. Reformatting outputs. Cross-checking the same tables. When AI absorbs that work reliably, the savings compound across the entire workflow. Even marginal improvements at multiple stages add up to something significant.

But it's important to frame this correctly. The value of AI isn't to replace the people doing the job — we still need people for the complex understanding, the contextual writing, the interpretation of what the data is actually telling us. What AI does is remove the repetitive effort, so that experts can focus on the higher-level, higher-value judgments that actually move things forward. In regulated environments, even small reductions in that kind of friction can translate into significant timeline improvements across an entire program.

5. You can't just drop AI into an existing process and expect it to work. What does genuine integration into an everyday operational workflow look like?

There's an organisational dimension to integration, and then there's the human one — and the second is often what determines whether an implementation actually sticks.

The people using these tools day-to-day are not tech professionals. They're domain experts under deadline and delivery pressure, working in environments where accuracy isn't optional but a must. They don't have the bandwidth to learn a complex new system, and they shouldn't have to. If the tool requires significant effort to understand or is poorly designed, it will be quietly set aside the moment things get busy — and in this industry, things are always busy.

When we were building TriloDocs, we deliberately drew inspiration from the interactions that already felt natural to our users. The way medical writers ask each other questions when they pick up a new protocol, for example. The instinct to search rather than navigate. The expectation that a system should surface what's relevant without requiring you to know exactly what to ask or prompt for. Design that starts from those moments of familiarity creates far less resistance than design that starts from capability.

The other piece is governance. Users need to trust that the system is operating within clear boundaries — that there are no surprises waiting for them when a senior reviewer or an inspector looks at the output. That confidence doesn't come from documentation. It comes from experience, and it builds over time when the tool consistently does what it says it will.

6. Most companies start with one AI tool. Why is that rarely enough — and how do you build from there?

Because most problems worth solving aren't one-dimensional. A single AI tool, however well chosen, is optimised for a specific type of task. The moment you move outside that scope — different data types, different outputs, different levels of precision required — you're asking it to do something it wasn't designed for.

In practice, a mature AI implementation looks more like an architecture than a product. You might have a rule-based system handling the deterministic processing, a language model managing narrative outputs, and a validation layer sitting across both. Each component does what it does well, and the orchestration between them is where the real value gets created.

We went through exactly this process with TriloDocs. The temptation to build one system that does everything is real — and it took discipline to resist it. The mistake is trying to build all of that at once. Start with the most clearly defined problem, implement something that solves it reliably, and let the architecture grow from there. Each layer you add should be connected to a specific need — not to a vision of what the system might eventually become. Organisations that take that incremental approach end up with something coherent and maintainable. Those that try to solve everything upfront tend to build systems that are expensive to run and difficult to explain.

7. How do you know which part of your business is actually ready for AI — and which parts aren't?

There are multiple signals worth looking for. The first is process clarity. AI performs best when it's operating against a well-understood workflow, where the inputs are defined, the business and expected domain logic is consistent, and the expected output is clear. If the process itself is still evolving or poorly documented, introducing AI into it will amplify the confusion rather than resolve it.

The second is repeatability. The strongest use cases are those where preset logic sets are being applied over and over, consuming time and capacity that could be better spent elsewhere. That's where AI delivers the most immediate and measurable return.

The third is data quality. Generative models in particular are only as reliable as what goes into them. Variable, inconsistent, or poorly structured data leads to outputs that require constant correction — which defeats the purpose entirely and creates significant downstream risk in regulated contexts.

Beyond these three, there's a dimension that gets overlooked in most readiness assessments: the people. Resistance to AI adoption often reflects legitimate concerns about job security, workload, and trust in the technology. Those concerns don't go away if you ignore them. Whether your teams genuinely believe the tool is there to help them — rather than replace them — will determine whether the implementation gains traction or stalls. And when choosing a vendor, it's worth asking whether they understand your user base well enough to support that conversation, or whether they're focused purely on the technical delivery.

8. For an operations or innovation leader starting this journey — what's the one thing you need to understand before they implement anything?

Simple: implementing AI is not just buying a tool. It's an organisational and architectural decision — a commitment to moving in a direction, not a one-time purchase.

Before anything else, understand your own organisation. What problem are you actually trying to solve? Who does this affect, and how ready are they? What operational overhead can you realistically absorb with the solution embedded? These questions matter because the answers shape everything — the tool you choose, the vendor relationship you need, and the timeline you set.

On that last point: think carefully about what kind of partnership you're looking for. AI is not a static solution — it evolves, and your needs will evolve with it. With TriloDocs, for example, the approach is deliberately non-invasive — minimal setup, with most of the operational overhead managed on our side. But some solutions may require significant internal investment to get running. Neither is inherently wrong, but you need to know what you're taking on before you commit.

And through all of it, keep the human dimension front and centre. You can have the right architecture, the right governance, the right tool, and still lose if the people using it don't want to. Implementation is as much a change management challenge as a technical one. Teams need to understand that this is there to make their work better, not to add five more projects on top with the time it saves on one. That message needs to be deliberate and consistent.

It's not an easy journey, and it shouldn't be treated as one. The organisations that navigate it well tend to be those that invest as much in the relationship with their implementation partner as in the technology itself — because the hard questions don't stop once the contract is signed. They're just getting started, and so are you.

Gaspar Wong is Head of Product at TriloDocs, where he leads the development of the company's AI platform for regulatory documentation. He brings a background in product leadership and cloud strategy, with experience delivering large-scale technology programmes across global organisations.

Lean Medical Writing: A Review of Principles, Applications, and Future Directions

Abstract

Lean medical writing represents a paradigm shift from traditional storage-based documentation to story-driven, interpretive communication in regulatory and scientific contexts. This review presents established principles of lean medical writing and examines their application in medical and scientific writing, particularly within pharmaceutical regulatory submissions. We explore the foundational concepts, evidence supporting lean approaches, and practical implementation strategies for Clinical Study Reports (CSRs) and other regulatory documents.

Introduction

The modern regulatory environment faces unprecedented information overload. Between 2003 and 2013, the FDA received 342 new molecular entity (NME) applications, with individual electronic common technical document (eCTD) submissions reaching up to 10 GB.¹ Simultaneously, scientific publishing has exploded, with over 2,000,000 papers published annually across approximately 20,000 journals, and cardiovascular research publications alone increasing by 14% annually.¹ This proliferation of information has created a challenge regarding communicating complex medical and scientific data efficiently without sacrificing accuracy or comprehensiveness.

Lean medical writing emerged as a solution to this challenge, drawing inspiration from lean manufacturing principles pioneered by Toyota and adapted for healthcare communication. Unlike traditional redundant medical writing, which is characterised by extensive copying of tabular data into text, lean writing emphasises interpretation and context. This review examines the principles of lean writing, evidence supporting its adoption, and practical strategies for implementation.

Defining Lean Medical Writing

The Storage vs. Story Framework

Medical writing exists on a spectrum from "storage" to "story." Traditional verbose approaches, described as "storage-based" medical writing, function like an Amazon warehouse where writers copy numerical values from tables into narrative text without adding interpretive value. This approach generates voluminous documents that provide little analytical value beyond the underlying data tables.

On the other hand, an opposite extreme approach exists. Very minimalist medical writing reduces text to brief statements such as "the data is in the table below." While concise, this approach forces reviewers to independently analyse all data without guidance, potentially missing critical patterns or requiring extensive review time. Lean medical writing occupies the optimal middle ground, providing interpretation while maintaining conciseness. This approach tells reviewers what the data mean, not merely what they are.

Core Principles of Lean Writing

1. Elimination of Wordiness

The foundation of lean writing involves removing unnecessary words, phrases, and repetitions.^2,3 Correia (2024) provides a clear example: instead of stating "The patient experienced a significant reduction in pain levels as a result of the administered medication," lean writing would say "The medication significantly reduced the patient's pain." Another example demonstrates this principle: instead of "There were no deaths in either treatment group during any phase of the study," lean writing simply states "There were no deaths."

This principle extends beyond individual sentences to entire sections. Writers should start by eliminating junk words — modifiers like "it seems," "it may," "might be," "could be" — followed by adjectives and adverbs that add no substantive meaning.³ The goal is not terseness for its own sake, but rather precision that respects the reader's time and cognitive resources.

2. Active Voice and Direct Expression

Lean writing prioritises active sentence structures for directness and engagement.² Rather than passive constructions like "The medication was prescribed by the doctor," lean writing states directly "The doctor prescribed the medication."² This approach clarifies agency and reduces word count while improving readability.

3. Logical Organisation

Information should be structured to guide readers efficiently through complex medical concepts, research findings, or treatment guidelines.² When discussing treatment options, for example, Correia (2024) recommends progressing logically: "The treatment options for this condition include lifestyle changes, medication, surgery, and alternative therapies," starting with initial interventions and progressing to more advanced options. This organisational principle has particular relevance in regulatory documents. Demographic data appears first in CSRs not by convention, but because if the medical writer doesn't know whether the treatment groups are comparable, they can't say anything about efficacy and safety. A lean approach makes this logical progression explicit.

4. Avoidance of Unnecessary Jargon

While medical writing necessarily employs technical terminology, lean principles emphasise using simple, commonly understood terms when appropriate.² Instead of "myocardial infarction," for general audiences, writers should state "heart attack," or instead of "dyspnoea," writers should state "breathlessness."² For specialised audiences, technical precision remains important, but jargon should never substitute for clear explanation.

Lean writing principles complement plain language approaches, particularly in patient-facing communications where accessibility is paramount.² The combination of lean and plain language techniques results in enhanced clarity for diverse audiences, including patients.² When medical content employs these principles, it facilitates efficient translation and adaptation into different languages and cultural contexts, ensuring content remains culturally appropriate and easily understood across populations.²

5. Emphasis on Clarity and Unambiguity

Messages must be clear and unambiguous to their intended audience.² Rather than stating "Based on the findings of the research study, it can be concluded that there is a positive correlation between regular exercise and cardiovascular health," lean writing would state: "The research study found a positive correlation between regular exercise and cardiovascular health."²

The Case for Lean Medical Writing

Regulatory Reviewer Perspectives

Regulatory reviewers have expressed clear preferences for interpretive documentation. At a Drug Information Association (DIA) meeting in Washington, FDA reviewers noted that high-quality CSRs are uncommon, with many documents presenting extensive historical background before reaching clinically relevant information. Reviewers indicated that CSRs often redundantly present tabular data in text form without added interpretation, diminishing their utility for regulatory assessment.

These perspectives reveal a critical gap between current documentation practices and reviewer needs. Regulatory authorities seek contextualized interpretation rather than redundant data presentation.

The Problem with "Well Balanced"

A specific example illustrates the inadequacy of minimalist writing that lacks interpretation. Many CSRs state that "treatment groups were well balanced" for demographics without further explanation. This phrasing is problematic because "well balanced" lacks standardised definition. For example, a difference of 5% may be considered balanced by some reviewers but not others, requiring each reviewer to independently assess the data.

The lean alternative explicitly addresses interpretation: "There were no differences between the treatment groups for demographic or baseline characteristics that would affect the interpretation of efficacy or safety," followed by explicit identification of any differences that might impact interpretation. This approach saves reviewers hours of independent analysis while maintaining transparency.

Efficiency Gains and Resource Implications

The business case for lean medical writing encompasses multiple stakeholders. Bhardwaj et al. (2017) note that "both writers and reviewers of these documents are overwhelmed with data and information overload," particularly when adhering to guidelines that inadvertently increase document volume. Given regulatory agencies, research organisations, and scientific journals operate with finite expert resources for document development and review, volume reduction becomes essential.¹

Latino (2024) emphasises that lean authoring in CSR development delivers cost efficiency by "focusing on essential information, ensuring CSR documents are straightforward yet thorough" while achieving "timeline reduction" through "reducing redundancy and implementing clear cross-referencing." The result is a win-win-win scenario: benefits for pharmaceutical companies (time and cost savings), reviewers (faster, easier reviews), and ultimately patients (faster drug approvals).

The Challenge of Measuring Impact

While the benefits of lean writing appear intuitive and align with reviewer feedback, quantifying specific improvements remains challenging. Lean writing can reduce review time and improve reviewer satisfaction, but it is important to remember that it cannot guarantee regulatory approval, which ultimately depends on the drug's safety and efficacy profile. Rigorous comparative studies would require identical compounds evaluated by identical regulatory authorities under equivalent conditions, making controlled research impractical.

Nevertheless, logical inference suggests that documents better aligned with reviewer preferences should facilitate faster, more favourable reviews. The FDA's general preference for lean, interpretive documents provides strong qualitative support for this approach.

Lean Writing in Practice

Process-Level Implementation

Bhardwaj et al. (2017) outline a systematic approach to implementing lean principles, drawing on Toyota's lean methodology. The process involves five cyclical steps:

1. Identify the deliverable (e.g., CSR or manuscript) as specified by customers
2. Examine the process and mark steps for elimination that don't affect quality
3. Develop documents according to the new workflow
4. Evaluate quality and performance based on customer feedback
5. Seek perfection and update the process iteratively

This framework emphasises that lean writing requires attention to both process and content. The goal is not simply to shorten documents, but to fundamentally rethink how information is analysed, synthesised, and presented.

Lean Authoring Techniques for CSRs

Latino (2024) identifies specific strategies for implementing lean authoring in CSR development. Firstly, rather than repeating data, strategic cross-referencing allows writers to reference other sections or source documents, which reduces document length and helps readers focus on critical insights. However, cross-referencing must be strategic, directing readers to protocols or statistical analysis plans rather than reproducing entire sections in the CSR. Latino (2024) recommends emphasising "essential findings over exhaustive data tables, keeping regulatory reviewers engaged and informed without overwhelming them." This approach aligns with the principle of story over storage, telling reviewers what matters rather than forcing them to discover it independently. Focused data presentation using summarised tables and cross-references helps avoid overwhelming stakeholders with data repetition. The goal is not data elimination but data contextualization, providing sufficient information while highlighting the most clinically relevant findings.

Challenges and Considerations

The "Mean" Component: When to Be Selective

Bhardwaj et al. (2017) introduce the concept of mean writing, emphasising strategic selectivity in content inclusion. They note that exhaustive data presentation, comprehensive analyses covering all possible angles, and redundant publication of methods impose significant burdens on all stakeholders — authors, reviewers, publishers, and readers alike.

However, they caution against eliminating exploratory and post-hoc analyses entirely. Such analyses, while not preplanned, can reveal subgroup efficacy or selectivity and inform decisions on inclusion and exclusion criteria.¹ The solution is clearly demarcating key, exploratory, and post-hoc analyses, with exploratory analyses presented as "short reports that supplement/complement the publications from the main study."¹

Resistance to Change

Despite widespread acceptance of lean principles in other industries, adoption in medical writing remains incomplete. Cultural resistance may stem from several sources. Medical writers often receive traditional training emphasising comprehensiveness over conciseness, creating ingrained habits that resist change. Concerns about omitting critical information lead to hesitation about adopting shorter document formats. Despite informal reviewer preferences for lean writing, the absence of explicit regulatory guidance creates uncertainty that inhibits adoption. Additionally, established organisational templates and processes create inertia, making systematic changes difficult to implement even when benefits are recognised.

As Bhardwaj et al. (2017) observe: "The concept of lean and mean medical writing is yet evolving and is yet to be globally accepted by the pharmaceutical companies, academics, and scientific communities." Successful implementation depends on coordinated efforts across pharmaceutical companies, regulatory authorities, academic institutions, and scientific communities, facilitated by detailed guidelines tailored to specific document types.¹

Importantly, as advocates work to promote lean writing adoption, the core message must remain clear: lean writing is not about minimalism for its own sake. The James Lind Institute (2012) emphasises that powerful medical writing contains vivid details and relevant examples that should not be eliminated. The goal is eliminating waste — redundancy, unclear expression, unnecessary jargon — while preserving and even emphasising essential narratives that provide context and meaning.

Technology and the Future of Lean Medical Writing

Electronic Formats and the Future of Lean Writing

Electronic submission formats enable new approaches to lean writing through strategic hyperlinking. Rather than repeating methodological details available in protocols, CSRs can hyperlink to specific protocol sections, reducing redundancy while maintaining accessibility.¹ However, this approach has yet to achieve full acceptance, as stakeholders often still expect complete materials and methods sections despite protocol appendices containing identical information.

Looking forward, Bhardwaj et al. (2017) envision a future where data transparency practices enable publication of clinical trial results with key highlights for patients and medical professionals, while detailed methodologies and raw data remain accessible through web links to publicly disclosed protocols and trial results. This vision represents the ultimate expression of lean principles: primary data openly accessible, with publications focused purely on interpretation.

AI and Symbolic Systems

While not extensively covered in the reviewed literature, the emergence of AI tools for medical writing raises important questions about lean principles. In an era of information abundance, the challenge has shifted from data availability to finding the needle in the haystack — identifying clinically meaningful signals within vast datasets. Generative AI systems that primarily copy and paste without interpretation risk perpetuating storage-based medical writing at scale. For instance, a generative AI tool might convert all demographic table values into text statements such as "The mean age was 45.3 years in the treatment group and 44.8 years in the control group. The mean BMI was 27.2 kg/m² in the treatment group and 27.5 kg/m² in the control group" — merely reproducing tabular data without interpretive value.

In contrast, systems using symbolic AI combined with interpretive capabilities may better support lean writing by consistently applying interpretive frameworks to data analysis. These systems operate on pre-defined rules rather than probabilistic decision-making. For example, a symbolic AI system might analyse the same demographic data and generate: "Treatment groups showed no clinically meaningful differences in baseline demographics (see Table X for detailed values), indicating groups were well-matched for efficacy and safety interpretation." The system applies consistent, pre-defined criteria for identifying meaningful differences — for instance, a clinically meaningful age difference of more than 5 years or BMI difference of more than 2 kg/m² will be defined and the table will be referenced rather than reproducing all values. This approach applies consistent criteria for identifying meaningful differences and provides context for why these comparisons matter, embodying lean writing principles through automated interpretation rather than mere data transfer.

Recommendations for Implementation

Organisations seeking to implement lean medical writing should consider the following strategic actions:

1. Develop Clear Standards: Establish document-type-specific guidelines that specify which data require interpretation versus tabular presentation, define criteria for clinically meaningful differences, and provide cross-referencing protocols.¹
2. Build Writer Capabilities: Train medical writers in interpretive analysis, emphasising the distinction between clinical and statistical significance, and developing judgment about meaningful differences.
3. Engage Regulatory Expertise: Attend regulatory forums, consult former FDA/EMA reviewers now in private practice, and contribute to industry working groups developing lean writing standards.
4. Implement Iteratively: Begin with pilot projects, gather feedback from internal and external reviewers, refine processes based on experience, and gradually expand as organisational comfort grows.¹
5. Track Meaningful Metrics: Monitor internal review cycle times, quality control finding rates, reviewer feedback quality, and stakeholder satisfaction to demonstrate value and build organisational support.

Conclusion

Lean medical writing represents more than a stylistic preference, as it embodies a fundamental shift in how we conceptualize regulatory and scientific communication. Moving from storage to story, from comprehensive data presentation to focused interpretation, lean writing aligns with regulatory reviewer needs, accelerates development timelines, and ultimately serves patients through faster access to effective therapies.

While challenges remain, particularly around standardisation, new challenges are emerging that require proactive attention. For example, as technology evolves, AI tools present both opportunities and risks for lean writing principles. Generative AI systems that merely reproduce tabular data risk perpetuating storage-based approaches at scale, while symbolic AI systems applying consistent interpretive frameworks may advance lean principles through rule-based analysis that prioritises interpretation over data transfer. The critical distinction lies not in automation itself, but in whether technology supports interpretation and contextualization — the hallmarks of lean writing. By implementing symbolic AI systems that combine pre-defined interpretive rules with analytical capabilities, organisations can scale lean writing practices while ensuring consistent application of clinical judgment across all documentation.

Effective medical writing translates complexity into understanding and meaningful healthcare communication.² In an era of unprecedented information volume, this translation function has never been more critical. The future of medical writing is lean, and organisations that embrace these principles position themselves for meaningful contribution to improving patient outcomes through efficient, effective drug development.

Dr. Barry Drees is the Founder and Advisor at TriloDocs, former President of the European Medical Writers Association (EMWA), and has spent 40 years in medical writing watching the field evolve from hand-constructed tables to AI-enhanced documentation.

What Big Pharma Is Really Telling Us About AI in Medical Writing

Over the past six months, conversations with senior medical writing and regulatory leaders across large pharmaceutical organisations have begun to converge around a clear and consistent message. Despite the rapid rise of Generative AI, enthusiasm inside regulated environments is becoming more measured. What is emerging is not resistance to AI, but a more mature, risk-aware understanding of where different technologies genuinely belong.

This shift closely mirrors the direction set out in the January 2026 FDA–EMA Guiding Principles of Good AI Practice in Drug Development, which emphasise human-centric design, risk-based use, clear context of use, and reliable, traceable outputs across the drug-development lifecycle.

In short: the conversation has moved on from whether AI should be used to how it can be used without breaking trust.

GenAI fatigue is driven by QC burden, not fear of innovation

Large pharma teams have actively explored GenAI for clinical and regulatory documentation. Many have run pilots. Some have deployed tools. Almost all report the same underlying problem: the QC burden is unsustainable. Senior leaders are increasingly explicit. Editing narrative for clarity or interpretation is acceptable. Verifying whether numbers are correct is not.

In practice, teams and recent evaluations report that off-the-shelf LLM-driven approaches deliver inconsistent performance in regulated pharmaceutical documentation tasks, such as clinical trial protocol generation. For example, assessments of leading models have shown strong results in areas such as content relevance, suitability, and appropriate use of terminology, but they fall short in critical clinical thinking and logic, resulting in outputs that are not 100% accurate and demonstrate inconsistent reasoning.^1-3 While such levels may be tolerable in low-risk domains, they are unacceptable for regulated documentation, where patient safety, inspection readiness, and traceability are non-negotiable.

This is why interest is shifting towards approaches that generate output directly from source data, behave deterministically, and eliminate the need to QC numbers altogether. This distinction aligns directly with the FDA–EMA emphasis on risk-based validation proportional to context of use, particularly where AI contributes to evidence generation. The issue is not a lack of innovation, but rather who bears the risk when the outputs are incorrect.

Lean medical writing is non-negotiable and poorly served by most tools

Another strong and recurring signal is dissatisfaction with the shape of AI-generated content. Many tools produce fluent, verbose outputs that look impressive but fail in practice. Senior leaders are clear that regulatory reviewers do not want bulk.

Across large organisations, internal discussions increasingly converge on the same conclusion. Lean medical writing is not optional. It requires:

• Minimal viable content
• Factual, data-driven interpretation
• Clear identification of clinically relevant signals
• Elimination of descriptive padding and narrative "fluff"

This challenge is amplified in organisations where therapeutic areas have historically written very differently. When verbose styles are imposed broadly or amplified by AI, the result is inconsistency, reviewer frustration, and documents that do not support decision-making. AI systems that generate more text rather than better judgement are increasingly seen as creating work, not removing it.

Deterministic approaches better reflect regulated reality

Importantly, this shift does not represent a rejection of GenAI altogether. Most senior leaders expect multiple AI technologies to coexist. What is changing is the consensus that high-risk, data-driven regulatory documents require deterministic behaviour.

Approaches that generate output directly from structured source data, apply consistent expert-defined logic, regenerate identically when inputs change, and clearly separate deterministic processing from generative assistance are viewed as fundamentally better aligned with regulated workflows than probabilistic, text-first drafting tools.

This reflects real operational experience. It also closely mirrors regulatory expectations around context of use, data governance, traceability, and lifecycle management for AI systems used in drug development. What is essential to understand is that not all AI belongs in the same risk class.

Human expertise must be amplified, not replaced

A further theme emerging from senior leadership is frustration with the assumption that AI should "replace" medical writers. In practice, the opposite is true. The most effective use of AI is where it removes mechanical, error-prone work, stabilises outputs across teams and vendors, and surfaces clinically relevant signals consistently — leaving human experts to do what only they can do: apply judgement, interpret meaning, and shape the scientific story.

AI that ignores this human–technology relationship quickly loses credibility, regardless of how advanced the underlying model may be.

Trust, usability, and predictability matter more than novelty

One of the most telling insights from senior leaders is that the underlying technology matters less than whether it can be trusted and easily applied. What consistently matters is a single coherent user interface, clear understanding of what the system is doing, visibility into which outputs are deterministic versus generative, and confidence that outputs can be relied on without hidden risk. AI that requires users to second-guess, verify, or reverse-engineer its outputs rapidly erodes trust, no matter how innovative it claims to be.

A clear convergence is emerging

Taken together, the message from large pharma leadership is remarkably consistent, and it is increasingly echoed by regulators themselves. Quality control of numerical data remains a non-negotiable requirement. Current accuracy levels are simply not acceptable in regulated documentation. Lean medical writing is viewed as essential rather than optional. Deterministic, risk-based approaches are seen as far better suited to high-risk content. Above all, trust, transparency, and predictability are considered more important than technical novelty.

This signals a shift away from experimentation for its own sake and towards AI systems that can scale responsibly inside regulated environments without increasing risk. The relevant consideration is not the inevitability of AI use in medical writing, but rather the selection of AI approaches that demonstrably meet the standards of trust and reliability required in regulated medical documentation.

A New Year Reflection: What Vikings, Zeppelins, and 40 Years of Medical Writing Taught Me…

As we start a new year, I've been thinking about how often we accept things as true simply because that's what we've always believed.

Let me share a few stories that have stuck with me and what they have to do with medical writing, AI, and the future we're building.

The Smoking Room on the Hindenburg

Years ago, I visited Flughafen Friedrichshafen, one of the main zeppelin ports in Germany, for the first time. They've rebuilt the cabin from the Hindenburg (the famous zeppelin that crashed and burned in New Jersey) and you can actually have dinner there using the original menus.

Fascinating place. But here's what struck me most: the Hindenburg had a smoking room.

Think about that for a moment. You're sitting under a giant balloon filled with hydrogen — the most explosive, flammable substance known to mankind — and someone decided: "You know what we need? A smoking room."

When I learned this, my first reaction was disbelief. But then I realised: cultural norms can sometimes lead us to make decisions that seem completely illogical in hindsight. At the time, smoking was so ingrained in society that the idea of not having a smoking room was probably more unthinkable than the safety risk it posed.

Vikings Never Wore Horned Helmets

Here's another one: when you think of Vikings, what's the first image that comes to mind? Horned helmets, right?

Except Vikings never actually wore horned helmets. That image comes from 19th-century romanticism and opera costumes, not historical fact. But the myth is so deeply embedded in our collective imagination that it's nearly impossible to dislodge. We "know" what Vikings looked like, except we don't.

What Our Eyes Tell Us vs. What We Know

Does the sun go around the Earth, or does the Earth go around the sun?

We all accept that the Earth goes around the sun because brilliant scientists proved it centuries ago. But what do our eyes show us every single day? The sun is clearly going around the world.

Here's another demonstration: draw two parallel lines of exactly the same length. Now add arrow tails — one set pointing outward, one set pointing inward. Your brain will immediately perceive one line as longer than the other, even though you just drew them the same length and you know they're identical.

The Müller-Lyer Illusion

You can measure them. You can prove they're the same. But your brain will still insist one is longer.

To me, that's a classic case of how the brain interprets things the way it wants, even when you know it can't possibly be true.

What This Has to Do with Medical Writing

Over the years, I've learned that our field has similar assumptions — things we accept as true because "that's how it's always been done." When I think on my career, I remember days, entire weeks, spent in my office going through computer printouts. Line by line. Page by page. Cross-checking data manually because that was the only way to do it. If I'm honest, I would probably get six or seven years of my life back if I could have avoided that work.

But here's the thing: at the time, we didn't question it. That's what the work required. Just like the Hindenburg needed a smoking room and Vikings wore horned helmets.

The Time to Be a Medical Writer

If I could choose any time to be a medical writer, I would want to be a starting medical writer now so I could use AI tools like TriloDocs to make my life easier and more interesting. Not because AI makes us obsolete, but because it eliminates the manual work that never required human judgment in the first place. All those days spent going through computer printouts? All of that would be saved with TriloDocs. That time could instead be spent on what medical writers actually do best, and that includes clinical interpretation, strategic thinking, understanding what data means for patients.

Questioning What We Accept

The lesson from smoking rooms, Viking helmets, and optical illusions is simple: just because something seems obviously true doesn't mean it is.

And just because we've always done something a certain way doesn't mean it's the best way — or even a sensible way. As we begin 2026, the conversations I've had with medical writers show a willingness to question assumptions:

• Should medical writers spend weeks manually processing data?
• Is speed the most important thing AI should deliver?
• Is the future of medical writing about automation, or about enhancing human expertise?

The doctors I spoke with this year reminded me of what matters: accuracy, transparency, trust. None of this black box, no hallucinations, none of that kind of business. What they valued most was something an experienced medical writer once told me: it provides peace of mind that the decisions I made reflect best practice.

Welcome to 2026

The Vikings never wore horned helmets. The sun doesn't orbit the Earth. And medical writers have better ways to spend their time than going through printouts for years. Sometimes the most obvious "truths" are the ones most worth questioning.

Here's to building a future in 2026 that's based on what's actually true, not just what we've always believed.

Happy New Year to everyone in our community.

—Barry

Dr. Barry Drees is the Founder and Advisor at TriloDocs, former President of the European Medical Writers Association (EMWA), and has spent 40 years in medical writing watching the field evolve from hand-constructed tables to AI-enhanced documentation.

TriloDocs - Key Advancements Over the Last 12 Months

Positioning for Regulatory, Medical Writing, Clinical, Biometrics, and Innovation Teams

Over the past year, TriloDocs has evolved from a specialised CSR engine into a mature, deterministic clinical regulatory-document platform designed for enterprise-scale accuracy, auditability, and writing velocity. These advancements directly support the priorities of Medical Writing, Regulatory Affairs, Clinical Operations, Biometrics/Statistical teams, and Innovation groups evaluating structured automation.

1) Deterministic Data Processing at Enterprise Scale

Matrix-based table parsing

TriloDocs now deterministically processes all clinical tables — including PK, PD, efficacy, safety, labs, KM plots, deviations, dispositions, demographics, medical history, and medications — with line-level fidelity. This eliminates hallucination risk and ensures consistent interpretation across all study arms and phases.

Flexible rule-based engine

Medical writers can now adjust critical signal-identification parameters (for example, cut-offs) to align outputs with protocol intent, statistical guidance, or sponsor-specific conventions.

Automated data mapping

Before generation, the platform identifies and displays all source tables to be used. Users can confirm or modify the mapping, ensuring full transparency and reducing downstream QC cycles.

2) Built-In Quality Control and Auditability

Traffic-light QC system

Section-by-section signalling with deep-dives into every table allows rapid troubleshooting and early-stage quality control.

Version history and regeneration

All generated versions can be viewed or downloaded, ensuring full audit trails and consistent regeneration when tables or analyses change.

Signal identification

The symbolic AI engine identifies clinically relevant signals deterministically, mirroring senior medical-writer reasoning while maintaining traceability and reproducibility.

3) Expanded Coverage Across Real-World Writing Workloads

Symbolic engine for data-heavy sections

TriloDocs now handles all sections that typically dominate writing time and QC — especially PK/PD, labs and data-heavy narratives that most tools cannot analyse and make comparisons to identify clinically relevant signals.

In-house GenAI medical-writing agents

For outputs not handled by the matrix parser, TriloDocs applies a guided GenAI layer built with medical writing guidance, incorporating analysis-specific rules to produce controlled narrative suggestions without introducing drift.

Lay Summary generation

A purpose-built, compliant AI module produces patient-friendly lay summaries from CSR content with multiple validating layers, ensuring regulator-aligned structure and enhanced readability.

4) Improved User Experience and Enterprise Readiness

New front-end platform

The interface has been fully modernised for clarity, speed, and multi-team value generation across Medical Writing, Regulatory, Clinical, Innovations and Biometrics.

Licensing model aligned to enterprise workflows

Sponsors retain template and data control while allowing external writing partners to work within project boundaries using the same deterministic engine.

5) Template Flexibility and TransCelerate Alignment

Use your organisation's TransCelerate template

TriloDocs supports direct use of your organisation's TransCelerate-aligned templates, preserving your established structure, styles, and compliance requirements. The deterministic engine adapts to your preferred format — whether you follow the TransCelerate standard exactly or customise it for internal conventions. The document is generated and downloadable in MS Word Docx format, ensuring seamless integration with your existing workflows and consistent document generation across all studies and vendors.

6) Upcoming Modules

• Investigator's Brochure — Spring 2026
• Module 2.7.4 (Summary of Clinical Safety) — Summer 2026

These additions expand TriloDocs into a broader structured-authoring ecosystem for regulatory documentation.

Why Global Teams Are Taking a Demo

Global pharma organisations are prioritising deterministic automation because it delivers measurable value across multiple functions:

• Predictable DBL→CSR timelines
• Reduced QC cycles and fewer late-stage escalations
• Improved inspection readiness and traceability
• Consistent outputs across internal teams and FSPs
• Scalable automation aligned with digital/AI transformation strategies

TriloDocs integrates symbolic AI, expert-authored rules, and guided GenAI into a single, compliant platform that generates inspection-ready documents in hours.

A short demonstration provides the most direct way to understand how this architecture can support your organisation's clinical and regulatory workflows. Use the "Book a Demo" button on this site and our team will reach out for a personalised demonstration.

Fresh look, same mission: TriloDocs unveils their new logo!

A logo may be small, but it carries a lot of meaning. Today, we're proud to reveal the new TriloDocs logo, an evolution of our identity and a reflection of our values: Accuracy. Efficiency. Innovation.

Why the change?

Because just like the field of AI, we're always growing. Our technology has become even more powerful, our mission has become sharper, and our vision for the future of clinical document creation is clearer than ever. We wanted a visual identity that captures that growth while staying true to our foundation: accuracy you can trust.

What makes TriloDocs different?

Our approach is built on three key pillars that set us apart:

1. Medical writing expert-level-developed rules

Our system is powered by rules crafted and validated by seasoned medical writers. This means we don't just summarise data, we also interpret it with clinical insight. From complex sections like PK/PD and Labs to subtle data nuances, our rules ensure accuracy and relevance even in the most challenging parts of a report.

2. Unique, matrix-based data parsing engine

Tables are processed deterministically, meaning there are no large language models applied. This way we ensure that data pulled through is always consistent and reliable. This rule-based approach removes the need for user data validation, because we get it right the first time.

3. Most accurate & consistent

By applying predefined rules to extract analytes and generate clinical observations, TriloDocs eliminates inconsistencies from the outset. This dramatically reduces QC time and ensures every table, graph, and data point is complete, correct, and traceable back to its source. Our system even detects subtle discrepancies that human reviewers might overlook.

This isn't just a new design, it's a symbol of progress. A step forward in showing who we are and where TriloDocs is heading.

What does this new look say to you?

Dr. Barry Drees Returns to TriloDocs as an Advisor

We're proud to share that Dr. Barry Drees, the founder of TriloDocs and the brilliant mind behind our rule-based AI — which applies expert medical writing knowledge to ensure unmatched data accuracy — has rejoined TriloDocs as an Advisor.

Barry's return comes at a pivotal moment as we expand and refine our product offerings post-integration into the Genactis family. His guidance will be instrumental in ensuring that TriloDocs remains the most accurate and reliable medical writing AI solution in the global market.

With a PhD in Molecular Genetics from the University of California and postdoctoral research funded by the U.S. NIH, Barry brings deep scientific insight and decades of industry experience to the table. He is a former EMWA President and past Editor-in-Chief of the EMWA journal, and continues to be a respected thought leader in the field.

We're thrilled to have him back and helping to shape what's next for TriloDocs.

TriloDocs Joins Genactis Group

TriloDocs is proud to announce that we have recently been acquired by Genactis, a global leader in healthcare research and technology.

This strategic move marks a significant milestone in TriloDocs' growth journey, positioning the company to scale innovation and deepen its commitment to transforming the way Medical Writing teams draft and deliver CSRs.

With Genactis' resources and global reach, TriloDocs is accelerating the product development roadmap, expanding support capabilities, and enhancing overall user experience, all while continuing to provide the same high-quality service and solutions our clients have come to expect.

"We are delighted to have secured TriloDocs and look forward to building out the product and customer base."
— Elgin Loane, Chair of Genactis

TriloDocs Users Can Expect:

• A New User Interface — cleaner, faster, easier to navigate
• API Connections — for seamless integration into your workflows
• Improved Usability — intuitive, streamlined experience
• New Features — based on real client use cases
• More Document Types — matching the diversity of your content needs

The future of TriloDocs is brighter than ever. Backed by the global expertise and resources of Genactis, we're entering a bold new chapter: defined by rapid innovation, deeper client collaboration, and focus on delivering meaningful value. Together, TriloDocs and Genactis are poised to redefine what's possible in healthcare content solutions.

We're excited for what's ahead — and even more excited to build it with you!