The Fundamentals of Omnichannel Still Apply: Q&A with Pat McGloin

The pharma and biotech industries are asking the same questions about AI as most other industries: what can it do, and what can it actually do? Over the past two years, the technology has built up a lot of hype over the possibilities.

Now, as people across industries become more experienced with it, they have a better understanding of AI’s abilities. They also understand its limitations much better.

Pharmaceutical Executive spoke with Pat McGloin, chief client officer and life science practice leader at MERGE, about AI’s potential in creating personalized HCP and patient engagement at scale. According to him, generative AI likely isn’t the sole answer for creating these systems while also ensuring regulatory compliance.

Pharmaceutical Executive: What are the dangers of AI hype for pharma?
Pat McGloin: The biggest danger I see is companies feeling pressure from leadership to “get into AI” before they have decided what problems they are trying to solve, what foundation the technology requires, or how it fits into the brand’s overarching strategy. That can create a lot of noise and activity without generating real progress.

Within pharma, I believe AI should operate inside a well-governed, walled-garden environment when possible. By that, I mean a secure and controlled system built on trusted data, approved content, clear access rules and appropriate regulatory oversight. That is especially important in an industry dealing with sensitive health information, complex treatment decisions and strict promotional requirements.

The lesson itself is not new. The same fundamentals we have spent years trying to get right in omnichannel still apply. Start with clean, organized and connected data. Establish a consistent taxonomy. Define what success actually looks like. Then reverse engineer the content, technology, measurement and operating model required to achieve it. Only after that should you determine where AI can amplify the system through greater speed, efficiency, insight or personalization.

The old adage still holds true: garbage in, garbage out. AI can simply process the garbage faster and distribute it at greater scale. If your data is fragmented, your content is disconnected or your objectives are unclear, AI will not rescue the strategy. It will magnify those weaknesses.

The opportunity with AI is real, but AI should support the commercial strategy, not become the strategy. Master the brilliant basics first. Then apply AI where it can make a well-designed system smarter, faster and more responsive without compromising trust, compliance or the human experience.

PE: In what ways is generative AI not capable of creating meaningful personalization?
McGloin: Generative AI can make personalization more precise, but it cannot make it meaningful on its own. At MERGE, we are strong proponents of human-in-the-loop AI. We see it as an executional enabler of connection, but not the source of the human insight that makes that connection feel real.

By now, we have all been exposed to AI slop: the endless stream of content that looks fine at first glance but, on closer inspection, reads like word salad and is instantly forgettable. Patients, caregivers and HCPs do not want or need more of that. No one does, really. People are more strapped for time and attention than ever, while the volume of information competing for that attention continues to grow. If we use generative AI simply to create more content faster, we are adding noise to a system that is already overwhelmed by it.

Where I see AI more broadly adding real value is in helping us identify and interpret signals that can make an interaction more relevant. Within an appropriately consented patient-support environment, for example, AI might recognize that a patient missed a refill, repeatedly visited safety information or stopped engaging with educational content. Those signals can help inform the message, timing and channel. But they do not necessarily tell us why. The patient could be concerned about side effects, struggling with cost, confused by the treatment plan or simply overwhelmed by everything else happening in their life. That context matters.

This is what we mean by Whole Human Marketing. Humans bring the strategy, empathy, creativity and judgment needed to create content that demonstrates an appreciation for the individual and their circumstance so we can better connect. AI can help shape and scale the creation process, identify relevant audience signals, and improve the targeting and delivery of that content so it reaches someone with information that is useful at a moment when it can actually help.

The goal is not to demonstrate how much we know about someone. That’s just creepy. It is to make the interaction feel useful, respectful, empathetic and uniquely human. In pharma, we’re often interacting with people who are navigating conditions that are profoundly personal. AI can help us make that connection at scale, but humans have to create the feeling behind it.

PE: How are leading pharma companies balancing AI innovation with MLR requirements?
McGloin: Leading pharma companies are using AI to strengthen the MLR process, not route around it. This is one of the areas where I think AI can genuinely advance what we are doing commercially by improving accuracy, reinforcing consistency and identifying potential issues earlier, while keeping final review and approval in human hands.

Modular content is a good example. It has long promised faster, more relevant content at scale, but it can quickly overwhelm an MLR process when every possible combination creates a new review burden. And MLR is rigorous for good reason. The stakes in our industry are high, and the work has to be thorough, truthful, accurate and appropriately balanced.

We have recently been piloting what I would call MLR enablement engines. A company can provide the system with pre-approved message maps, content matrices, claims, references, disclosures and rules defining how those elements may be used together. AI can then assemble candidate variants within those boundaries, include the associated references and disclosures, and flag deviations before the content reaches human review.

That does not mean every new combination is automatically compliant or approved. Context still matters. A claim that is accurate on its own can become misleading depending on its placement, audience or the content surrounding it. The final decision still belongs to MLR.

AI can also provide a first level of quality control within the natural flow of work. It can compare creative assets and generated content with approved claims, supporting references, brand standards and configured regulatory requirements. It can flag issues such as missing fair balance, unsupported language, outdated references or inconsistent versions earlier in the process, so reviewers spend their time on content that is actually ready for them.

The real value is not lowering the standard or removing people from the process. It is reducing preventable errors, repetitive manual checks and unnecessary rounds of review. That allows medical, legal and regulatory experts to spend more of their time on the decisions that genuinely require context, expertise and human judgment.

PE: What are some practical strategies for scaling AI while maintaining compliance?
McGloin: Scaling AI responsibly requires turning compliance from a final checkpoint into part of the operating system. The principle is similar to manufacturing best practices: you cannot inspect quality into a product at the end of the line. You have to embed a quality mindset and the right controls throughout the process.

The first step is recognizing that not every AI use case carries the same level of risk. An internal tool summarizing approved research is very different from a system generating patient-facing content or recommending a next-best action. Classify each use case by data sensitivity, audience, autonomy and potential harm, then match governance and review requirements to that risk.

Keep AI activity inside secure, governed environments. Use approved models, minimize the data being shared, apply role-based access and establish clear retention rules. Prompts, outputs, sources and model versions should be logged to create an audit trail. Sensitive health, customer or company information should never end up in a public model simply because it is convenient.

The rules also need to show up in the tools people actually use. Static PDFs have a role, but they should not be the primary way employees encounter the rules. Put approved terminology, claims, source content, prompt templates and escalation rules directly into the tools teams are using. Custom assistants can be grounded in approved material, require source attribution, flag outputs that move outside established boundaries and route higher-risk work to MLR.

Every use case should also have a named owner, defined approval and override authority, routine output audits and a process for responding when a model, source or regulation changes. “Human in the loop” cannot mean someone simply clicks “approve” at the end. Start narrow and earn the right to scale by proving that a use case improves accuracy, consistency or efficiency without creating unacceptable risk. If the foundation works, expand it. If it does not, fix it so you’re not automating the problem at scale.

PE: What should commercial leaders prioritize if AI adoption continues?
McGloin: Commercial leaders should use AI to make human interactions more valuable. Simply automating more activity is a pretty low bar. That means moving beyond disconnected pilots and building an operating model in which data, content and human insight continuously inform one another.

Field sales is a practical example. Today, many reps work from a call list and compete for limited time with HCPs so they can share information that supports informed prescribing. But rep-led interactions and non-personal promotion often operate in separate systems. The rep sees one part of the HCP journey, while digital channels see another.

By connecting those interactions within appropriate privacy and compliance guardrails, AI can create an intelligence feedback loop. Learnings from rep and HCP conversations can help improve future content and outreach, while engagement signals from non-personal channels can help reps prepare more effectively. We are using AI to identify hand-raisers whose engagement suggests they may be looking for additional information. AI can surface that signal and help prioritize outreach, while the rep determines whether and how to act on it.

AI can also help focus the conversation by identifying topics that appear most relevant based on the signals we are receiving. That does not mean scripting the interaction or assuming we know exactly what an HCP needs. The signals provide a better starting point. The rep still brings the expertise, curiosity and judgment required to understand the individual and make the conversation genuinely useful.

This is what I mean by an infinite intelligence system: each interaction creates feedback that can make the next interaction smarter and more relevant. Operationalizing this approach requires clean, connected data, a shared taxonomy, clear ownership and strong governance. Done right, this creates fewer, better and more meaningful human connections.