SPEAKERS:
Tom Baker, SVP, Global Data, Technology, and Advisory Services, IQVIA (Moderator)
Diana Medeiros Plácido, Chief Digital & Technology Officer, ViiV Healthcare
Chetak Buaria, VP, Global Commercial Operations, Merck
KEY TAKEAWAYS:
· AI deployed within functional silos produces faster functions, not faster organizations.
· Operating model redesign must precede AI deployment; sequencing determines whether AI integrates or entrenches.
· Regulatory and behavioral silos require different interventions; incentive misalignment is the deepest and least-addressed layer.
· One-third of physicians already use YouTube or LLMs daily for clinical updates, making current engagement architectures structurally obsolete.
· AI at functional handoff points is the design principle that distinguishes silo-breaking from silo-scaling.
Pharma companies are deploying AI to break down organizational silos. The tool is making them worse. Diana Medeiros Plácido, Chief Digital & Technology Officer at ViiV Healthcare, put it plainly: "AI is great in augmenting. Its already used widely to augment roles but If you're not careful, AI will augment your silos and not break your silos. And this is an important perspective, right? Because if suddenly you continue with the same way of working and thinking and you're just focusing on a particular function, AI will augment your silo." Tom Baker, SVP, Global Data, Technology, and Advisory Services at IQVIA, moderated the session at Pharma 2026. He had opened with the premise that silos carry legitimate purpose while generating the drag that undermines competitive speed: "Silos are an obvious enemy of speed. We hear this repeatedly. The organizations that we operate in tend to have silos and they're often there for a reason. Where does the friction in the system that's been built up over time, where does it show up most visibly?"
Plácido's warning reframed everything that followed. The question was no longer how to deploy AI against silos, but whether the organizational structures receiving AI were designed to benefit from it, or simply to scale existing dysfunction at machine speed.
The Operating Model Must Precede the Technology
The value of cross‑functional coherence in pharma is well established. Plácido referenced research showing that prescribers who experience a strong customer experience are more than twice as likely to prescribe, controlling for safety and efficacy. In other words, what differentiates products in competitive markets is not just science, but how consistently the organization shows up across commercial, medical and digital interactions. That is an organizational design problem before it is a technology problem.
Yet the dominant AI deployment pattern in pharma runs in the opposite direction. A commercial next best action engine built here. A medical insight platform standing up there. Each tool optimized for its function's KPIs, measured by its function's leadership, governed by its function's data. Baker stated what that produces: "If you just deploy the AI in the function or the silo, you've now just reinforced the silo. Maybe a greater efficiency, but it's still a silo."
Increasingly, pharma needs to rethink tools like next‑best action not as a sales optimization tool, but as a cross‑functional engagement capability. If next‑best action is built to serve a single function, it will optimize that function — and reinforce the silo around it. If it is built as a cross‑functional engagement capability, it can enable the organization to coordinate actions across roles and reduce handoffs in the moments that matter most.
This is why AI so often disappoints at scale. The technology is the same, but the outcomes aren’t. AI deployed inside a silo accelerates siloed behaviour. AI deployed into a model with shared ownership, shared metrics and shared data across functions begins to change how the organization works and drives Customer experience.
The Incentive Layer Beneath the Org Chart
Distinguishing between types of silos changes the available interventions entirely. Chetak Buaria, VP of Global Commercial Operations at Merck, drew the foundational line between structures that must be preserved and behaviors that must be changed. Regulatory compliance frameworks, data privacy requirements, and patient safety guardrails are what he calls "hard silos" — legitimate, non-negotiable, and appropriately restrictive. The target is the "soft silos": behavioral habits, insight attrition at handoffs, and information hoarding that occurs when functions have no structural reason to share. But Buaria placed a ceiling on what technology can fix: "Silos will continue to exist as long as humans are in the loop. It's just the mindset thing, and that's what it boils down to."
Below the behavioral layer sits something harder still. Buaria identified incentive architecture as the variable most organizations address last, if at all: "Are we incentivizing our R&D organization or regulatory affairs to get the regulatory approval, or are we incentivizing ourselves to make maximum impact of the drug that reaches to the maximum number of patients as fast as we can? So it also is sometimes the undercurrent that needs to be addressed up front." When R&D success is defined by milestone achievement and commercial success is defined by patient access, no shared dashboard reconciles the misalignment. Each function is rational. The organization is incoherent.
Buaria made this concrete with a clinical distinction that carries strategic weight: "It's one thing to take a drug to regulatory approval, but it's completely different thing to see if that approved drug is eventually reaching and making a difference to the patient. And that's where the challenges lie." Fast to market and fast to patient impact are not the same goal. Organizations that treat them as equivalent discover the gap in the field, when a drug that cleared every approval milestone fails to make an impact in the market.
His answer to the question of personal orientation was disarmingly direct: "I stand up to make sure that my HCPs have right information at right time and patients who are deserving the drug get them as quickly as possible. So that's my title that matters more than the functional identity that I carry in my organization." Every internal title, every functional KPI, every performance review cycle is invisible to the patient. The organization that recognizes this redesigns incentives accordingly. The one that doesn't builds cross-functional task forces that dissolve at the first budget cycle.
The discussion also highlighted a less visible but decisive factor: accountability. While operating models, data platforms, and incentives set the conditions for speed, progress stalls without clear ownership of cross‑functional outcomes. Organizations that move faster tend to establish explicit decision rights, often supported by trusted internal or external partners, accountable to enterprise‑level outcomes such as time to patient impact. In this framing, governance is not about controlling risk, but about enabling faster, coordinated decisions across silos.
Discover more on this topic at Pharma Commercial Data & Tech Europe 2026 (4-5 November, London) Europe's collaborative home for data and tech pioneers. Visit the website here.
When the Promotional Target Disappears
The operating model work described above is urgent. It may also be racing against a more disruptive structural shift.
Baker pointed to behavioral evidence that the current commercial model is already eroding: "We did some research last year. We saw that fully one third of physicians use YouTube at least once a day to stay up to date on clinical and scientific information. And so that's an entirely different customer profile and the experiences that they're looking for." Younger physicians are significantly more likely to use large language models and show declining interest in face-to-face engagement, a generational shift in information consumption, not a temporary preference adjustment.
Buaria named the more radical scenario emerging from his oncology practice: "We might end up living in a scenario where a lot of those choices that physicians are making today will be taken over by agentic solutions at their end. And how from a promotion mix perspective, you position as an industry in that kind of constellation is a bit scary as well because it completely disrupts the current way of marketing drugs." Complex hospital systems deploying their own AI agents to mediate treatment decisions would not simply change which channels pharma uses to reach physicians, it would change the role that the physician plays.
The operational response Plácido described is instructive precisely because it shows the silo problem replicating itself in real time around every new channel. Gen AI search optimization is already triggering fragmented, independently siloed responses across functions such as commercial, medical, and communications functions: "Suddenly commercial is talking about it, medical is talking about it...How are you going to bring it back into an organizational strategy and aspirations, to reevaluate how we work as a result of an evolving world? What are the guardrails?" Every new engagement surface creates a new coordination challenge before it creates a competitive advantage. Organizations that have not resolved the underlying operating model question will face this cycle indefinitely, each new channel spawning a new silo before the last one has been addressed.
AI Works at the Handoff, Not Inside the Function
The paradox has a resolution, but it requires a specific architectural choice: deploying AI at the interfaces between functions rather than within them.
Buaria provided the clearest proof from his experience of deploying AI at the functional interface of Medical, Legal and Regulatory (MLR) review process: "We started seeing already the massive improvement in the workload with help of embedding Gen AI in MLR which was trained on all the comments that were made for last two years on our materials...it already gives me five different suggestions that will help me reduce my review cycle time and to improve my chances [of MLR approval]." The MLR use case works because it targets a handoff, the intersection of marketing, medical, legal, and regulatory review, rather than making any single function faster in isolation. The model doesn't accelerate marketers or reviewers independently; it compresses the friction between them. The architecture of the deployment is the insight.
Baker identified the structural lock-in that makes this harder than it appears: "Functions exist for a reason...those functions are also generating their own data products...And those then are sort of perpetuated by the functions." Proprietary data products create self-reinforcing information architectures that resist integration even when the strategic intent is explicitly cross-functional. The technology layer inherits the organizational layer. Building a unified customer data platform on top of six functional data products doesn't dissolve the underlying ownership logic; it buries it.
Plácido placed the binding constraint where it belongs: culture.
“Culture eats strategy for breakfast,” she noted. “You need to build an organisation that is curious, willing to experiment, and not afraid to change the way it works — because the environment is changing, and it will continue to change.”
That cultural shift isn’t an abstract ambition. In pharma, it is central to staying anchored on patient needs in an environment that is becoming more complex and less linear. When organizations lack digital fluency or are fearful of experimentation, they default to safe, siloed behaviours. Digital fluency and a culture willing to learn and adapt are therefore prerequisites for patient‑centred impact. Adoption is not the starting point; it is the outcome of an organisation that is designed — and culturally equipped — to respond faster and more coherently to what patients actually need.
Pharma's silo problem operates on three distinct layers; structural (functions and org charts), informational (separate data products and insight flows), and motivational (divergent incentives rewarding contradictory outcomes). Current AI deployments address the informational layer almost exclusively. The structural layer gets attention during reorganizations. The motivational layer, where functions are measured against outcomes that do not and sometimes cannot align, receives the least intervention and does the most damage. Technology redesigns information flow. Only incentive redesign changes whether people use it.
The diagnostic for any pharma leadership team is straightforward: audit your current AI portfolio. If every deployment lives within a single function and is measured by that function's KPIs, you are not breaking silos. You are building faster ones.
All authors are employed by pharmaceutical companies or by entities providing services to the pharmaceutical industry. The views and opinions expressed in this publication are solely those of the authors and do not necessarily reflect the views, positions, or policies of their respective employers.
To get you highlights of Pharma 2026 faster, we are using generative AI technology to summarise the transcripts of the sessions. If you have any feedback about the summary, please contact lucy.fisher@thomsonreuters.com.
Discover more on this topic at Pharma Commercial Data & Tech Europe 2026 (4-5 November, London) Europe’s collaborative home for data and tech pioneers. Visit the website here.
