Breaking silos in healthcare: how AI is enabling more personalised care

By Dr Marzena Nieroda (UCL Global Business School for Health) and Prof Philip Treleaven (UCL Computer Science). 

Healthcare is becoming more personal. We increasingly hear terms such as personalised care or hyper-personalisation, especially in conversations about artificial intelligence (AI). These concepts are well developed in e-commerce and service industries, where they are at the cutting edge. In the healthcare context, however, for many people they remain unclear, over-promised, or disconnected from everyday healthcare experiences.

This blog draws on recent research at UCL, specifically two research papers—one on the Open Health platform and one on how people interact with AI through prompts—to explain, in plain language, what personalised and person-centred care really requires. The work comes from an interdisciplinary collaboration across UCL Global Business School for Health and UCL Computer Science, led by Dr Marzena Nieroda and Professor , and reflects ongoing research within the Person-Centered Care Lab. Together, these papers explore how AI, data, governance, and system design can work together to support more personalised care—if we address a deeper challenge first.

That challenge is not a lack of data, technology, or expertise. It is a lack of connection.

The ideas discussed here are shared early as open preprints, reflecting how innovation now happens in fast-moving fields like AI and healthcare. Important design and policy decisions are often made before research appears in journals, and preprints allow these ideas to be discussed, tested, and improved while systems are still taking shape. We hope these papers will stimulate discussion, invite contribution from researchers, practitioners, and decision-makers, and help catalyse innovation in practice. Readers who would like to engage more deeply with the technical detail, evidence, and design trade-offs are warmly encouraged to explore the full papers, which are openly available and underpin the insights summarised below.

At its core, this research makes one simple argument: personalised, person-centred healthcare is not primarily an AI or knowledge problem—it is an integration problem. The rest of this blog explains why silos stand in the way of personalised care, how AI fits into this picture (with both promise and limits), and where new opportunities are emerging when we design systems to connect knowledge rather than keep it apart.

Why Personalised Healthcare Is So Difficult to Achieve

At a human level, personalised care sounds obvious. Everyone is different, so care should reflect individual needs. In reality, healthcare systems struggle to deliver this consistently.

The main reason is fragmentation. Health knowledge is spread across many places:

• doctors and hospitals focus on clinical records and test results,
• researchers work with selected datasets, often separated from daily care,
• public health bodies rely on population-level statistics,
• patients generate growing amounts of data through apps, wearables, and everyday life.

Each source captures something important, but rarely are they brought together. As a result, care can feel generic, research insights may not reach practice, and innovation often remains stuck in small pilots. From a person-centred perspective, the biggest barrier to personalisation is not a lack of knowledge—it is silos that prevent knowledge from being connected across disciplines.

What We Mean by Hyper-Personalisation

In everyday settings, personalisation usually means simple tailoring—such as a reminder or recommendation based on past behaviour. In healthcare, this approach quickly reaches its limits.

In our work, hyper-personalisation refers to care that:

• draws on many kinds of information at once,
• recognises that health changes over time, and
• supports different people involved in care, not just one user defined by the lens of disease.

In plain terms, hyper-personalisation is about joining the dots. It means understanding how medical information, lifestyle, environment, and social context interact—and using that understanding to support better decisions.

Crucially, this cannot be achieved by looking at one dataset or relying on AI alone. It requires systems that integrate knowledge across boundaries that currently keep health information apart.

AI: Powerful, Promising—and Still a Work in Progress

AI clearly has a role to play in this shift, but it is important to be realistic. AI systems are not perfect. They can miss context, oversimplify complex situations, or reflect the limits of the data and questions they are given.

In healthcare, these weaknesses matter. Poorly designed AI can reinforce silos, produce confident-sounding but incomplete answers, or fail to reflect individual circumstances. These challenges are real and should not be ignored.

At the same time, our research deliberately focuses on opportunities, not just flaws. The key question is not whether AI is good or bad, but how it is designed, governed, and used. When AI is built to support integration—across data, systems, and people—it can become a powerful enabler of more personalised, person-centred care. The papers explore these trade-offs in detail and explain where further development is needed.

The Open Health Idea: Making Shared Knowledge Possible

One of our papers introduces the Open Health platform, which addresses a core limitation of today’s systems: how organisations can learn together without moving or exposing sensitive data.

In simple terms, Open Health:

• lets data stay securely where it is,
• allows approved analyses to run locally, and
• shares only safe, agreed-upon results.

This technical approach matters because it enables knowledge sharing without data pooling. It allows medical, lifestyle, and environmental information to be connected in ways that respect privacy, trust, and local control.

From an innovation perspective, this is not just a technical upgrade—it is a new foundation for personalised, person-centred care. Readers interested in the architecture, demonstrator, and governance model can find full details in the Open Health preprint paper

Talking to AI: Why Questions Matter for Personalised Care

The second paper looks at how people interact with AI tools—such as chat-based systems—to explore health information.

These tools are becoming common entry points to knowledge. They are helpful because they make information more accessible. But they also depend heavily on how questions are asked.

If questions are too broad or too simple, important differences between people, situations, and timeframes can be lost. Our research shows that meaningful personalised care requires better conversations with AI—questions that reflect real-world complexity, uncertainty, and individual context.

This is not just a technical issue; it is a design and education challenge. The prompting paper provides concrete examples of why this matters and how interaction-aware approaches can improve outcomes.

Personalised Care Through Better Connections

The central message of this work is simple: more personalised, person-centred healthcare depends on better connections.

AI is not a silver bullet, and it still needs careful development. But when designed to connect knowledge rather than deepen silos, it opens new opportunities for care that is more relevant, transparent, and human.

For readers interested in understanding both the opportunities and the limitations in more depth, the full papers provide detailed explanations, examples, and evidence—and are strongly recommended as a next step.