Kelly H. Zou Speaker Interview

Over the last year, what has changed most significantly in how organizations think about the value of real-world evidence?

Answer:

When I look back over the past year, the most significant change in how organizations think about real‑world evidence (RWE) is the shift from seeing it as a promising complement to clinical trials to recognizing it as a strategic necessity. Recently, I have become a New York City (NYC) Ambassador for Anthropic’s Claude in both Healthcare and FinTech, as well as Chair, Caucus of Industry Representatives, and President, NYC Chapter. I also continue as the Chief Executive Officer of AI4Purpose to generate impact of artificial intelligence (AI) for social good.

As the field has matured, and as expectations from regulators and payers have continued to rise, this shift has become increasingly visible across research and development (R&D), medical affairs, and market access teams. In last year’s piece, I talked about how real‑world data (RWD) sources have expanded dramatically, from claims and registries to electronic health records, wearables, mobile applications, and even social media, and how this expansion has fundamentally changed the scale, diversity, and immediacy of the evidence we can generate. Taken together, these developments have pushed organizations to think more intentionally about how RWE fits into their broader evidence strategies, prompting deeper conversations about quality, governance, and decision‑readiness that align closely with real‑world evidence strategy.

Despite unprecedented investment in data, analytics, and AI, many organizations still struggle to turn evidence into decisions. Why does this gap continue to persist?

Answer:

Despite unprecedented investment in data infrastructure, advanced analytics, and artificial intelligence (AI), many organizations still struggle to turn evidence into decisions. And the reasons are more human than technical. In many cases, teams are overwhelmed by the sheer volume of data but lack the governance frameworks needed to ensure quality, provenance, and interoperability. Even when high‑quality analyses are produced, they may not reach decision‑makers at the right moment or in the right form. In my experience, the pace of analytics often outstrips the pace of organizational decision cycles. AI can generate insights in minutes, but committees may meet monthly, and governance processes can take weeks.

That mismatch creates friction. There is also a persistent siloing of evidence generation and decision‑making. Analysts may produce sophisticated models, but if the insights are not embedded into operational workflows, they remain underutilized. This is why I often emphasize the importance of integrated evidence planning and operational design. Evidence alone is not enough; organizations need pathways that carry evidence into decisions.

Many companies are investing heavily in integrated evidence planning. From your perspective, what separates organizations that successfully align evidence generation with payer needs from those that continue to face delays, restrictions, or unfavorable access decisions?

Answer: Successful organizations, in my view, ask themselves the kind of questions mentioned above, catalyzed by an outside-in way of thinking. This doesn't mean there are easy answers, but it can lead to augmented insights for external stakeholders.

AI has moved beyond experimentation and into real-world evidence workflows. What distinguishes organizations that are successfully scaling AI from those that remain stuck in pilot mode? 

Answer:

AI has now moved beyond experimentation and into real‑world evidence workflows, and the organizations that are successfully scaling AI share several distinguishing characteristics. My Clinical Leader article on probabilities of technical and regulatory success (PTRS) was ranked the top 5th, which illustrates this aspect. It explains how probability of technical success (PTS) and probability of regulatory success (PRS) combine to form PTRS, a metric that guides portfolio prioritization, investment decisions, and risk mitigation. AI enhances these metrics by analyzing historical clinical trials, regulatory precedents, and biological data to identify patterns that human analysts might miss.

However, those organizations that scale AI effectively do more than build models. They treat AI as an enterprise capability. They invest in shared platforms, common standards, and reusable components. They prioritize data quality and provenance, recognizing that AI is only as strong as the data it learns from. They embed AI into existing workflows rather than creating parallel processes. And they invest in change management, training, and communication so that teams understand how to interpret AI‑generated insights. In contrast, organizations stuck in pilot mode often have impressive demonstrations but lack the infrastructure, governance, and cultural readiness needed for scale. The difference is not the sophistication of the models; it is the maturity of the ecosystem around them. This is why exploring AI scaling patterns is important.

We hear a lot about "decision-grade evidence" today. What does that term actually mean to you, and how should organizations assess whether their evidence is genuinely decision-ready?

Answer:

The term regulatory-grade RWE has become increasingly common, but its meaning is often imprecise. To me, decision‑grade evidence is evidence that meets the methodological, operational, and contextual standards required to support a specific decision with confidence. It is not a universal label; it is decision‑specific. Evidence that is decision‑grade for a safety signal may not be decision‑grade for a regulatory submission or a payer negotiation. Organizations should assess decision‑readiness by evaluating data quality, methodological rigor, transparency, reproducibility, and alignment with regulatory or payer expectations. They should also consider whether the evidence is timely, interpretable, and actionable. Decision‑grade evidence is not just statistically valid; it is fit for purpose. It must be trustworthy, contextualized, and delivered in a form that supports the decision at hand. RWE is not something you declare at the end of the data analysis, but rather via integrated planning from the beginning.

How do you see the relationship between regulators, payers, and evidence generators evolving over the next few years, particularly as expectations around RWE continue to increase?

Answer:

The relationship between regulators, payers, and evidence generators is evolving rapidly as expectations around RWE continue to increase. Regulatory agencies such as the United States Food and Drug Administration (FDA), European Medicines Agency (EMA), and Pharmaceuticals and Medical Devices Agency (PMDA) in Japan are becoming more explicit in their guidance, offering clearer frameworks for how RWE can support approvals, label expansions, and post‑market commitments. Payers are demanding more rigorous evidence to support value‑based agreements, reimbursement decisions, and assessments of real‑world effectiveness. Evidence generators—whether biopharmaceutical companies, academic institutions, or technology partners—must navigate these rising expectations while ensuring that their methods, data sources, and governance structures meet the needs of diverse stakeholders.

Over the next few years, a greater convergence between regulatory and payer expectations has emerged, more collaboration on methodological standards, and increased emphasis on transparency and reproducibility. This evolution will require ongoing dialogue, shared learning, and a commitment to building evidence ecosystems that are robust, ethical, and patient‑centered. It is an area where deeper exploration of regulatory collaboration can be particularly helpful.

These days, I am particularly interested in exploring how RWE and AI are reshaping product development and evidence strategy from methodology to impact. One area I am eager to discuss is the integration of unstructured data, including clinical notes, imaging, digital biomarkers, and patient‑generated health data, into evidence generation workflows. These data sources hold enormous potential for improving patient understanding, refining phenotypes, and accelerating insights that were previously inaccessible. Yet they also introduce new challenges related to quality, bias, interoperability, and privacy. Exploring how organizations can responsibly harness unstructured data, supported by AI, will be a critical part of the conversation. It is a topic that sits at the intersection of science, technology, ethics, and patient experience, and it reflects the broader transformation underway in how we think about evidence.

Your keynote panel at IMPACCT is focused on how RWE and AI are shaping R&D and evidence strategy from methodology to impact. What's one area of discussion you're particularly looking forward to exploring with your fellow panelists?

Answer:

Looking ahead as the co-chair of IMPACCT RWE 2026 summit, I believe the meeting comes at exactly the right moment for the field. We are entering a period where evidence generation is becoming continuous, adaptive, and deeply integrated into an end-to-end strategy for a product lifecycle. Let’s all shape conversations that are both scientifically grounded and operationally realistic, especially around how organizations can responsibly harness unstructured data, strengthen governance, and build decision‑grade evidence frameworks that truly influence development, safety, access, and patient outcomes. The summit enables new partnerships, methodological standards, and ways of thinking about RWE and AI well into the near future.

*Disclaimer: The view in this article may not necessarily reflect the author’s company’s views. No editorial assistance was provided.