The Med Comms Bottleneck: Why AI is Targeting Pharma’s Communication Problem

Medical communications has always operated under pressure: biopharma companies generate enormous amounts of clinical data – trial results, real-world evidence, safety updates – that must reach multiple audiences simultaneously, including specialist physicians, community doctors, medical science liaisons, payers, caregivers and patients.

Each audience requires different framing, language, and levels of technical depth. For decades, however, the people responsible for bridging that gap – skilled scientific communicators at medical affairs agencies – have spent a surprising portion of their working hours not thinking, but reformatting.

Moving slide content from one congress template to another, rebuilding decks for different audiences, and doing it manually often against tight overnight deadlines. “We would do all these deliverables for clients, but oftentimes we also spent so much of our time pulling together the presentation and then transferring things from one template to the next template,” said Francine Carrick, a PhD-trained scientist who spent 22 years in med comms.

“We dreamt of a solution that would translate that science for us,” she added.

Carrick recently joined AI presentation platform Prezent as the president of Prezent Vivo, which fuses purpose-built AI and domain experts to power the life sciences communication ecosystem – including both biopharma and the company’s agency partners.

The problem she describes isn’t niche; it sits at the intersection of two pressures now well-documented in the industry. On the one hand, nearly 8 in 10 healthcare professionals receive a greater volume of information from pharma companies than before COVID-19, and 77% say the volume of digital communications is already too great.

On the other, pharma companies are struggling to deliver the personalized, relevant content that HCPs need, partly because legacy systems lack the flexibility to support advanced personalization at scale. The content pipeline is overwhelmed at both ends: too much being produced, and too little of it landing effectively.

The Modular Content Problem

The industry’s proposed solution to this has long been “modular content” – the idea of breaking scientific information into reusable components that can be assembled differently for different audiences.

In theory, it’s elegant, but in practice, large language models are now being used to prepare manuscripts, condense real-world evidence datasets into summaries, and develop modules for educating healthcare professionals – tools that until recently existed only as proof of concept.

Carrick frames the underlying challenge in straightforward terms: “The way that we present to an academic physician versus a community doctor versus a caregiver versus a patient is very, very different,” she stressed.

“In the traditional model, taking that information and customizing it was very labor intensive, and it took time.” In other words, the bottleneck wasn’t the expertise of communicators; it was throughput – more data arriving faster than teams can manually repackage it.

Following widespread AI experimentation in 2024, companies are under pressure to show real returns on their AI investments, driving adoption of vertical AI solutions purpose-built for specific workflows.

This is exactly the argument Prezent is making with its Astrid AI agent: that a system built specifically for life sciences, trained on the compliance requirements, regulatory constraints and scientific vocabulary of biopharma, will outperform a general-purpose tool retrofitted for the industry.

The Question of Specialty

Whether the life sciences context genuinely demands purpose-built AI, or whether it’s a marketing framing for a competitive market is a legitimate question.

What’s clear, however, is that the FDA has been paying close attention. Following its publications of guidelines in 2025 on the use of AI to support regulatory decision-making for drug and biological products, it had received over 500 submissions containing AI components. Such regulatory scrutiny creates a real argument for compliance-native AI tools rather than adapted ones: the risk of getting it wrong in a regulated environment is qualitatively different from getting it wrong in, say, in a marketing deck.

The broader healthcare AI market reflects rising confidence: the global healthcare AI market is expected to grow from $26.6 billion USD in 2024 to 187.7 billion by 2030, with the industry already deploying AI at more than twice the rate of the broader economy.

Within that, pharma and biotech companies remain the most R&D focused, with 54% prioritizing innovation and drug development, though commercial operations – including communications – are increasingly on the agenda.

The Human Expertise Question

The arrival of AI tools in professional services reliably generates the same conversation: what happens to the people who currently do this work? In med comms, where the work requires genuine scientific fluency, the answer is more nuanced than the displacement suggests.

Carrick’s view is that the binding constraint on human expertise in med comms is not knowledge, but rather bandwidth. “It enables, it accelerates, the human expertise to many degrees,” she said of AI in her field. “It enables that expertise, those insights, that human knowledge to be shared with more audiences in a more timely fashion.”

This take aligns with what’s emerging as a more textured picture of AI’s effects on skilled knowledge work. Surveys of physicians suggest resilience instead of displacement, as many believe AI will change their work but not eliminate their role.

The analogy to medical communications isn’t perfect, but the structural similarity holds: what AI can currently do is accelerate the routine; what it can’t do is replace the scientific judgement, the audience intuition, or the strategic thinking that defines the higher-value work.

An EPG Health study found that nearly 60% of pharma respondents identified HCP insight as the top priority for strategic engagement, and that medical science liaisons have overtaken the Salesforce as the most important channel for sending information to HCPs.

That’s a signal that pharma is moving towards more relationship-intensive, less broadcast-style communications, which requires more human judgement, not less – even as AI handles the production layer.

What the shift actually requires

The harder question isn’t whether AI will play a role in medical communications – that’s already settled. It’s whether the tools being built are genuinely fit for the complexity of the domain.

Carrick noted what Prezent calls “fingerprints” – audience-specific communication preferences that can be encoded and applied when adapting content. The concept reflects a more fundamental challenge: that the goal of “the right content, at the right time, through the right channel, to the right audience” – long a mantra in med comms – has always been more aspirational than operational.

Delivering on it requires not just good science writing, but systematic knowledge of how different audiences process different kinds of information.

Whether AI can reliably encode that knowledge, and whether it can do so while maintaining the scientific accuracy and regulatory compliance that distinguish medical communications from other content industries, remains the open question.

Regardless, what’s no longer in question is that the old model, with its overnight formatting marathons and manual template migrations, was never sustainable at the pace modern biopharma demands.

The problem was visible to practitioners for years, but the tools to address it are only now becoming viable – thanks to AI.