In episode five of Pharma Decrypted, Sabine Louët was joined by Emmet Kearney, CEO at Altuent, a company working in the knowledge management space. Kearney made the case that the biggest risk facing pharma content today isn’t the content you are creating now, but the content you created years ago and haven’t revisited.
Here are the key points from the conversation:
1. Knowledge management isn’t information management
Kearney opened with a definition of what knowledge management is, and why it isn’t the same as information management. Information management is handling documents, data, and content within an organisation. It’s a systems question: where things are stored, how they are versioned, who has access.
Knowledge management is getting the right information to the right person at the point of use, in a form they can trust. It covers explicit, implicit, and tacit knowledge, and Kearney made a solid note that it’s a very old discipline. Long before modern content management systems existed, organisations transferred knowledge through storytelling, by asking a colleague how something was done.
So what’s changed? The audience. And increasingly – the consumer of that knowledge is not only a person but an AI system. This shift raises the stakes when it comes to accuracy and governance.
2. Pharma’s structure makes this harder than it looks
The siloed nature of pharma organisations is a particular challenge. R&D and clinical teams research generate knowledge upstream. Medical affairs and commercial teams need to use it downstream. Making that flow work has always been a persistent problem.
Kearney was direct about the limits of the conventional fix: The narrative around breaking down silos and building a single source of truth has been circulating for years, but achieving it across an enterprise is a very difficult task. What has improved is the ability to bridge silos rather than dissolve them, using semantic layers to connect systems and provide the context that allows the right information to surface when it is needed.
But he pointed to a deeper problem that bridging doesn’t solve. What happens when five systems hold five different versions of the truth? Claims libraries illustrate the issue well. A claim is signed off in one vault, then disseminated across many channels. When that claim changes, the risk isn’t the change itself. The risk is the legacy content sitting across the organisation that no longer aligns with it.
3. Structured content solved a problem and created another
Over the past several years, pharma companies have adopted structured content approaches, breaking content down to component level so that a single approved source can feed multiple channels. That has been a successful and meaningful step forward for omnichannel delivery.
The new challenge? Keeping those components fresh at scale. An organisation might hold thousands of signed-off facts and claims. When something changes in one system, identifying every downstream component affected by that change, and resolving the resulting conflicts, has traditionally been an extremely manual process.
Kearney described the approach his team takes: an agent layer that sits across the different repositories, checking facts and flagging inconsistencies when they arise. If a piece of content created six months after another contains a subtle difference, the system flags that difference to whoever owns it.
4. The problem is rarely the technology
The question of how much digital maturity an organisation needs before these systems become viable, Kearney was clear that the technology is rarely the constraint.
What separates organisations that move past pilots is governance, a clear use case, and an honest assessment of the risks. He described coaching a client who wanted to switch on AI agents to retrieve internal information. The advice was to build the backend first: understand what structure the content needed, add semantic assets, put a taxonomy in place. Only then does turning the system on make sense.
He shared that a technology partner came to him with a client getting poor results from an AI agent. Kearney explained the approach: content structure, context, taxonomy. The partner stopped and asked whether you would not just build a piece of technology to handle it.
That reaction, Kearney suggested, captures the gap. He has taken to asking organisations building new AI functions whether they have a knowledge manager on the team. The answer is often no, despite the presence of numerous data engineers and AI specialists. The work of establishing context is unglamorous, but it’s fundamental to what makes these systems work at scale.
Louët noted that this mirrored SciencePOD’s own experience developing a personalised clinical news feed for healthcare professionals. Without a taxonomy mapping disease areas, indications, and therapy areas, there is no way to route a relevant piece of news to the oncologist, the breast cancer specialist, or the dermatologist. Context is what makes personalisation possible.
5. The business case for fixing stale content has changed
Perhaps the most significant shift Kearney described concerns the economics of content maintenance.
Ten years ago, the conversation about updating outdated content rarely went anywhere. His team would build content for a campaign, that content would age, and the business justification to go back and remediate it simply didn’t exist. The argument that outdated content erodes trust was true then, but it didn’t move budgets.
Today, the calculation is different. Content is no longer only read by people who navigate to a website. It is retrieved, summarised, and presented by AI systems that may draw on a source that is six months, a year, or three years out of date. The risk is that a patient or clinician typing a natural language query into a chatbot can receive an answer that is factually incorrect because it was pulled from content the company forgot it had published.
6. Search has changed, so the content strategy must too
The conversation closed on the implications for how pharma organisations structure and measure their content.
Traditionally, performance was measured in visits to websites and time spent on page. Increasingly today, the reader may never land on the website at all. They may be searching through a chatbot, or receiving a generative answer in a search engine result.
Kearney’s view was that good SEO practice remains valid, and the principles behind Google’s E-E-A-T framework still apply from a generative engine optimisation perspective.
But content structure matters more now than it did previously. FAQs and well-organised medical information pages support retrieval by AI systems in a way that unstructured prose does not.
Underpinning all of it is the same question that has always mattered: who is doing the searching, and what are they trying to find out? Understanding the persona and the use case comes first. The structure, the taxonomy, and the governance all follow from that.
The Key Takeaways
- Know the difference between information and knowledge management. Storing content well isn’t the same as ensuring the right person, or the right system, retrieves accurate information at the point of need.
- The risk sits in your legacy content. When a claim changes, the exposure is not the change itself. It’s every piece of misaligned content still live across your channels.
- Structured content needs a freshness mechanism. Component-level content enables omnichannel delivery, but thousands of signed-off components need a systematic way to stay current.
- Let the machine check and the human decide. AI agents can surface inconsistencies at scale. Only a subject matter expert can determine which version is correct and why.
- Governance beats technology. Organisations that move past AI pilots do so because they have a clear use case, a governing layer, and an understanding of the human’s role, not because they bought better tools.
- The business case has shifted. Outdated content that once looked like a minor housekeeping issue now carries real risk when AI systems can surface it as an authoritative answer.
- Structure your content for how people actually search. Readers may never reach your website. Content structure, taxonomy, and context determine whether your information is the one that gets retrieved.
The views expressed in this interview are the speaker’s own.
Interview conducted by Sabine Louët. Edited from the interview transcript for clarity and length.