Across Asia Pacific, ageing populations, rising chronic disease burdens and persistent workforce shortages are increasing pressure on healthcare systems to improve clinical decision-making, operational efficiency and care delivery. The region’s healthcare AI market, valued at approximately $4.6 billion in 2023 and projected to grow at a 40.9 per cent CAGR through 2030, reflects a broader shift from experimentation to scalable healthcare outcomes.

Google DeepMind recently launched a national AI partnership with the Singapore Government to apply frontier AI to real-world challenges, reinforcing the nation’s ambitions in healthcare and life sciences. From clinical decision support to biomedical research, the promise is clear: faster discovery, more personalised care and more efficient healthcare delivery.

While much of the conversation around AI focuses on increasingly sophisticated models, the limitation now is less about model capability and more about data readiness. Healthcare organisations are generating vast volumes of data, yet much of it remains fragmented across institutions, systems and formats. To move AI safely into production, they must revisit the foundation: governed hybrid data architectures that balance accessibility, performance and compliance in a highly regulated environment.

Frontier AI Introduces New and Complex Data Requirements

Emerging AI applications in healthcare, such as AI co-clinicians, biomedical research assistants, and agentic AI systems, significantly increase the complexity of data requirements compared to earlier use cases.

Healthcare data itself is highly diverse. It spans structured data such as electronic health records and laboratory results, unstructured information like clinical notes and diagnostic images, and real-time data generated by monitoring devices and connected health systems. Each data type introduces different governance and management challenges, making document intelligence a critical part of healthcare AI readiness. Before a model can generate reliable outputs, unstructured information must be accurately interpreted and converted into usable, governed inputs. If the data pipeline misreads a document, misses context or reconciles records incorrectly, even the most advanced model can produce a confident answer from a flawed input.

Healthcare organisations therefore need data infrastructure that can integrate these sources, preserve clinical context and apply governance consistently before data reaches the model.

Agentic AI systems add another layer of complexity. Unlike earlier tools that mainly supported search, summarisation or prediction, agentic AI can operate with greater autonomy across tasks and systems. This means requiring continuous and reliable access to governed data, alongside strong safeguards that define what they can access, recommend or act on.

Fragmented Healthcare Data Remains the Greatest Barrier to Safety Deployment

Healthcare data is inherently decentralized across hospitals, polyclinics, laboratories, and research institutions, and often stored across incompatible systems with varying formats, standards, and governance frameworks. For AI systems that depend on consistent, high-quality data, this fragmentation limits the ability to generate accurate, reliable, and actionable insights.

As healthcare organizations seek to scale AI in a region where the APAC healthcare analytics market is projected to exceed $41 billion by 2030, data readiness is now a strategic priority rather than a technical consideration.

The concern is not about the lack of data, but the inability to generate value from it. According to Cloudera’s Data Readiness Index 2026, healthcare organizations globally face significant data governance and infrastructure roadblocks, as 45 per cent of respondents do not have access to 100 per cent of the data needed for AI initiatives.  KPMG and APACMed found that 73 per cent of Asia-Pacific healthcare leaders cite that data silos significantly limit data utilization, while 97 per cent of hospital-generated data remains unused. As a result, AI systems often spend more time locating, integrating, and reconciling data than generating meaningful clinical, operational, or research outcomes.

Healthcare institutions must therefore establish governed, interoperable data foundations. This means creating common data standards, shared metadata, consistent governance policies and secure access controls that allow data to be used where it resides.

Beyond Centralization: Balancing AI Innovation with Privacy, Security, and Patient Trust

Fully centralized approaches become increasingly difficult to sustain as AI initiatives scale. Healthcare data already spans on-premises systems, cloud platforms, research environments, and partner networks. Consolidating all data into a single repository can increase operational complexity, risk, and cost.

A more effective approach is to design for interoperability. This allows an institution to replace one component without changing everything, or without forcing all its data into one single place or vendor. In practice, a hospital should be able to use one storage system, another analytics engine and a locally adapted AI model without copying sensitive patient data into a proprietary island.

Hybrid architectures support this approach by allowing healthcare organizations to securely access and analyze data where it resides, reducing unnecessary data movement while supporting AI deployment at scale.

Trusted Data Pipelines Build Confidence in AI Systems

Trust is especially critical in healthcare, particularly in clinical decision-making and research environments. Clinicians and researchers must be confident that AI-generated outputs are accurate, explainable, auditable, and compliant with regulatory requirements.

That confidence depends on the data pipeline behind the AI system. Organizations need visibility into where information originates, how it has been transformed, and how it is used throughout the AI lifecycle. Strong data governance practices provide the foundation for this: data lineage helps trace information from source to output, auditability allows decisions to be reviewed and validated when necessary, and robust access controls protect patient privacy while enabling appropriate data sharing among authorized stakeholders.

By mandating participation in the National Electronic Health Record across all licensed healthcare providers through the Health Information Bill, Singapore is setting a blueprint for trusted data sharing — improving data coordination while maintaining strict controls over access, cybersecurity and patient privacy.

Singapore's ambition to become a leader in AI-powered healthcare and life sciences is well within reach. Realizing this vision, however, will require more than advanced algorithms or larger models. It will depend on strong, governed data foundations.