AI Readiness in Pharma – Getting the Foundation Right
Contributed Commentary by Narasimha Kumar, Global Head of Technology and Data Services, BC Platforms
June 18, 2026 | Many years ago, when AI was seen more in science fiction than in the real world, we were asked a seemingly simple question in an innovation workshop – why is it that our bank accounts and credit cards can be used globally, but we can’t do the same with our health information? The participants spoke about the challenges of interoperability, and the need for patient-centric harmonization.
Recently, I was in conversation with a pharma executive, who wanted to know how quickly AI can help accelerate evidence generation.
“Before we talk about AI,” I said, “can we bring together your clinical, genomics, and real-world data? Can it legally and securely be accessed across borders?”
We often talk about AI readiness as if it were mainly about algorithms, cloud infrastructure, or choosing the right large language model. However, in life sciences, AI readiness is fundamentally about evidence (data) readiness.
Without addressing evidence readiness, the best AI stack cannot accelerate drug development and get therapies to patients faster.
From molecule discovery and protocol optimization to patient recruitment and digital twins for clinical studies, the pharma industry is not short of ambition. Yet the reality is stark: AI is only as good as the evidence ecosystem that feeds it.
To understand disease progression, predict treatment response, or identify the right cohorts for a trial, organizations increasingly need to combine multi-modal data: clinical trial data, electronic health records, claims, imaging, pathology, genomics, registries, biomarker data, and patient-generated information.
In theory, this sounds manageable. In practice, it can feel like assembling a thousand-piece puzzle where half the pieces belong to different manufacturers.
In conversations with colleagues, I compare AI in drug development to designing a city’s transport system. Imagine trying to plan roads, metro lines, and commuter routes without fully understanding where people live, where industries are concentrated, or how neighbourhoods connect. Then, just as construction begins, you discover underground water systems, sewage networks, and power infrastructure that no one has properly mapped. The engineering may be world-class, but without a complete understanding of the ecosystem, delays, redesigns, and unintended consequences become inevitable.
In many ways, building deep clinical evidence, particularly across countries with fragmented healthcare systems, evolving regulation, and uneven data access, feels similar.
The challenge becomes even more complex outside the US.
The United States, despite its own interoperability problems, still benefits from relatively large-scale healthcare datasets, and a more unified legal environment. But once you move into Europe, Asia-Pacific, the Middle East, or multi-country global studies, the complexity increases dramatically.
In many programs, the science itself is relatively straightforward, but the evidence strategy becomes a geo-legal logistical challenge. A study spanning Germany, the UK, and several EU markets may involve entirely different consent frameworks, local ethics requirements, data residency expectations, and varying interpretations of privacy legislation.
The truth is this: the patient journey may cross borders, but health data often cannot. This fragmentation slows down the creation of deep clinical evidence, which AI needs to generate meaningful insights.
Take rare diseases or oncology as an example. To build robust evidence, researchers often need to aggregate patient cohorts across multiple countries because no single geography has sufficient scale. Yet navigating country-level legislation, institutional policies, and access restrictions can add several months to evidence generation timelines.
Regulation exists for a good reason. Cybersecurity threats in healthcare are increasing, and clinical trial data carries enormous scientific and commercial value. This is precisely where the industry needs to rethink what AI readiness actually means. Too often, organizations think of AI readiness as procuring tools: a new GenAI platform, a machine learning initiative, or creating a new function for AI innovation.
The organizations making real progress invest in trusted evidence ecosystems. Increasingly, the winning approach is not “move all the data to the AI.” It is “move the analytics to the data.”
We are seeing encouraging examples of this mindset already. Secure research environments such as the Wessex SDE as part of the UK NHS Secure Data Environment Network and controlled-access models like Finnish Biobank Cooperative, demonstrate that valuable research can happen with high-quality data without compromising privacy or governance. Federated analytics, privacy-enhancing technologies, synthetic data for model prototyping, and Trusted Research Environments are increasingly proving their value.
It is not necessary that every organization should adopt the same model. It is that trust, security, and interoperability can no longer be afterthoughts. They have to be baked into the design of the evidence generation platform.
In my view, true AI readiness in pharma rests on five foundations.
First, evidence readiness: harmonized, longitudinal, multimodal data that can answer meaningful clinical questions. We need to ensure the causal AI models are trained with the right data.
Second, governance readiness: clear consent models, provenance tracking, and auditability to know exactly where data came from and how it is being used.
Third, cyber readiness: secure-by-design architectures capable of protecting sensitive patient and research data.
Fourth, interoperability readiness: common standards and ontologies that allow data to speak the same language across systems, institutions, and countries.
And fifth, operational readiness: multidisciplinary teams where clinicians, researchers, data scientists, compliance leaders, and security specialists work together rather than in silos.
None of this sounds as exciting as talking about the next frontier AI model. But in practice, these foundations determine whether AI delivers measurable clinical impact or another expensive experiment.
To shorten development timelines and bring therapies to patients faster, we need to stop treating AI readiness as a technology procurement exercise. The focus should be on building trusted, interoperable, privacy-preserving evidence ecosystems that allow AI to thrive.
To hark back to our transport system analogy, AI may be the engine that transforms drug development, but no city runs efficiently because of vehicles alone. It works because the roads, tunnels, utilities, and governance are designed to function together. Drug development is no different.
Narasimha Kumar brings 25+ years of experience in technology consulting, AI product management, and compliance across life sciences and healthcare. Previously chief product and commercial officer at Datafoundry AI and senior leader at PAREXEL International, Narasimha specialises in AI-driven platforms and cloud transformation. He can be reached at [email protected].