From record to intelligence: How EMR systems on AWS become the foundation for generative AI in healthcare

Sooner or later, we all become patients. When that moment comes—for us, for someone we love—we don’t think about data formats or infrastructure architectures. We think about being seen. Being understood. Being cared for by someone who knows our full story, not just the last chapter.

While healthcare providers strive to deliver personalized and insightful care for every patient, they also operate in an environment of compounding data complexity. Electronic Medical Record (EMR) systems were designed to capture and store clinical data, but they weren’t built for the AI era. The core tensions are:

Data fragmentation: Clinical data is scattered across EMR systems, genomic and medical imaging data stores, clinical notes, metabolic data, clinical trials, and regulatory records, each with its own ontology, data stores, and processing frameworks.
Interoperability debt: Proprietary data models lock clinical information inside vendor silos, making it difficult for AI agents to access the contextual knowledge that guides human clinical decision-making.
Data access: Better healthcare delivery depends on the availability of all data for diagnosis and treatment. The same is true for AI: agentic success is directly constrained by data readiness, because agents lack the implicit contextual knowledge that clinicians rely on.
Modernization pressure: Regulatory mandates (especially in Europe: General Data Protection Regulation (GDPR), European Health Data Space (EHDS), and national frameworks in Germany and France) are forcing hospitals and HealthTech solution providers to rethink how clinical data is stored, governed, and shared.

With a clear mandate for healthcare transformation and innovation driven by payers, government, providers, and patients, these tensions highlight the gaps in existing systems’ architecture and capability. Healthcare providers are operating under inherited constraints created by current systems that lack basic data sharing and integration. And, primary among these systems is the EMR.

The two paths to unlocking clinical intelligence

For healthcare organizations and technology partners building on top of EMR systems, there are two complementary paths to unlock the full value of clinical data. The most forward-thinking organizations are pursuing both.

Strategy 1: Extend the existing EMR with AWS AI services

The first path works with what already exists. Clinical data remains in the EMR and AI capabilities are layered on top using foundation model access and agentic orchestration for multi-agent clinical workflows, natural language processing to extract clinical entities from unstructured notes, and custom model training for population-specific patterns using:

Amazon Bedrock for foundation model access and agentic orchestration (including Amazon Bedrock AgentCore for multi-agent workflows)
Amazon Comprehend Medical for extracting clinical entities from unstructured notes
Amazon SageMaker for custom model training and ML pipelines
AWS Glue and Amazon Simple Storage Service (Amazon S3) as the data foundation layer, with semantic enrichment using knowledge graphs (Amazon Neptune) and vector stores (Amazon Bedrock Knowledge Bases)

This approach delivers fast time-to-value for independent software vendors (ISVs) already embedded in a specific EMR ecosystem, and for hospitals seeking to unlock AI capabilities without replacing their existing systems. The data stays in the proprietary store; AI capabilities are layered on top through APIs and event-driven architectures.

Strategy 2: Integrate with open standards-based clinical data repositories

The second path goes deeper. Instead of layering AI on top of proprietary systems, it builds on open clinical data models: FHIR R4 for interoperability and openEHR archetypes for clinical meaning. AWS HealthLake provides the managed FHIR R4 datastore that underpins this strategy, enabling healthcare organizations to store, query, and enrich clinical data at scale. Together, they give AI the precision it needs to reason correctly across systems, languages, and borders.

This is the architecture that the EHDS regulation is mandating. It enables AI applications that aren’t tied to a single EMR vendor, and is particularly powerful for multi-provider environments and national health networks.

Ecosystem partners including vitagroup, Better.care, and Greenway Health are already building on these open standards with AWS as the cloud backbone for compute, storage, governance and AI.

This strategy is particularly relevant for multi-provider environments, national health networks, and organizations preparing for EHDS compliance. It enables AI applications that aren’t tied to a single EMR vendor.

These two strategies aren’t mutually exclusive. The most sophisticated organizations will pursue both in parallel.

Max Grundig Klinik: From decision to diagnosis in a single morning

At Max Grundig Klinik, a private clinic in the Black Forest region of Germany, the adoption of HeartFlow FFRCT illustrates exactly what cloud-centered healthcare infrastructure makes possible.

HeartFlow FFRCT uses CT imaging data to create a personalized, three-dimensional model of the coronary arteries, computing fractional flow reserve values non-invasively. For patients, this means a clinically validated alternative to invasive cardiac catheterization; reducing procedural risk, patient stress, and unnecessary interventional workups.

What would have taken days or weeks to deploy on-premises was live within a single morning. No hardware procurement, no server configuration, no extended IT project—only a secure, cloud-delivered service ready for clinical use.

Integration with the clinic’s existing Hospital Information System (HIS) and Picture Archiving and Communication System (PACS) was seamless. CT imaging data is automatically transferred from the PACS to HeartFlow through DICOM, analysis results flow back directly into the HIS, and the entire diagnostic workflow runs from image acquisition to clinical report without manual intervention.

“HeartFlow is a perfect example of how cloud infrastructure changes the equation for a clinic like ours. We went from decision to go-live in one morning. On-premises, that would have been a project lasting weeks. The cloud doesn’t just make us faster—it makes us more capable of bringing the right diagnostic tools to our patients at the right time.” – Arne Lehmann, CIO, Max Grundig Klinik

The Max Grundig Klinik example demonstrates what becomes possible when infrastructure is no longer the bottleneck. But speed of deployment is only one dimension: for clinical AI to scale across systems, providers, and borders, the data itself must speak a common language.

AWS and open standards: A deliberate commitment

AWS has made a deliberate and public commitment to open healthcare data standards, positioning the platform as the neutral, standards-compliant backbone for clinical data interoperability.

AWS HealthLake and FHIR

AWS HealthLake is a fully managed, HIPAA-eligible and part of AWS’s ISO- and SOC compliance programs. It accelerates healthcare interoperability by transforming fragmented healthcare data into a unified repository with performance at petabyte scale. Key capabilities include:

A high-performance FHIR R4 infrastructure with enterprise-grade security and sub-second latency
Built-in natural language processing to extract clinical context from unstructured medical text
Zero-ETL data access at petabyte scale for analytics and AI workloads
Support for AI-powered patient profiles in combination with Amazon Bedrock. For more information, see AI-Powered Patient Profiles using AWS HealthLake and Amazon Bedrock

AWS Partners like DataArt are already using HealthLake to deliver FHIR and EHDS accelerators, achieving measurable outcomes such as 50% reduction in prior authorization turnaround times. For more information, see Building trust in healthcare data with AWS HealthLake and DataArt.

AWS and openEHR

AWS is a Diamond Member of the openEHR International organization, demonstrating a strategic commitment to the openEHR standard as a vendor-neutral clinical data architecture. openEHR archetypes provide the semantic precision that AI models need to reason correctly about clinical concepts across different systems and languages.

“For decades, we have stored clinical information without truly defining it. Open standards like openEHR change that by making each clinical concept explicit, interoperable, and reusable, regardless of the system that records it or the vendor that hosts it. Cloud platforms then provide the operational answer: how to scale, secure, and surface that data inside the clinical moment. AWS’s Diamond membership of openEHR International shows a serious commitment to building that foundation in the open data space, and it is the kind of partnership that healthcare AI needs if it is to be trustworthy at scale.” – Jordi Piera, CEO OpenEHR

Bridging to EHDS

The European Health Data Space (EHDS) regulation mandates that member states enable both primary use (direct care) and secondary use (research, AI, public health) of health data, with FHIR and openEHR as the preferred technical standards. AWS investments in HealthLake, openEHR membership, and EHDS-aligned reference architectures (including eDelivery-compliant messaging and SML/SMP protocol support) position AWS as the preferred cloud platform for EHDS compliance across Germany, France, and the broader EU.

With the data foundation in place, the question shifts. It’s no longer about how to structure clinical data. It’s about what intelligent systems can do with it.

Amazon Connect Health – When the data is ready, intelligence follows

A properly structured, semantically enriched, cloud-centered EMR does something remarkable: it transforms a record of the past into a resource for the future. When clinical data is governed, interoperable, and clinically meaningful, it becomes more than a repository. It becomes the foundation on which intelligent agents can reason, recommend, and act.

Amazon Connect Health is the AWS answer to a specific question: what does agentic AI look like when it’s built for the people who deliver care? Amazon Connect Health is a purpose-built solution that handles high-volume administrative tasks—including appointment scheduling, clinical documentation, and medical coding—directly within existing EHR workflows. It builds on Amazon Connect Customer, the customer experience service that handles over 16 million interactions daily and extends it with healthcare-specific AI capabilities.

For healthcare provider organizations

Amazon Connect Health offers pre-integrated patient engagement agents:

Patient verification (generally available): Conversational patient identity confirmation with real-time EHR integration, eliminating manual record lookup at check-in
Appointment management (preview): Natural language voice scheduling with all day availability, real-time insurance verification, and EHR integration, so patients can book, reschedule, or cancel without hold times

For EHR companies and healthcare ISVs

A unified SDK provides three point-of-care capabilities that healthcare technology builders can adopt incrementally:

Patient insights (preview): Reads structured and unstructured FHIR data from AWS HealthLake and produces visit-specific summaries, including patient background, health events since last visit, and Hierarchical Condition Category (HCC) recapture opportunities
Ambient documentation (generally available): Captures patient-clinician conversation in real time and generates structured SOAP notes formatted to existing EHR templates, supporting more than 22 clinical specialties
Medical coding (preview): Generates ICD-10 and CPT codes from clinical notes with confidence scores and full source traceability

Amazon Connect Health can read directly from AWS HealthLake. Every patient insight, clinical note, and billing code traces back to its source transcript or patient chart data. Clinicians can view the underlying evidence for any AI output immediately. Patient-facing agents automatically escalate to a human when situations require it.

Healthcare organizations including Amazon One Medical, Netsmart, Veradigm, Greenway Health, and Pelago are already using Amazon Connect Health.

The numbers that should change everything

The architecture described in the preceding sections isn’t theoretical. Health systems across Europe, the UK, and Australasia have already made this transition, and the outcomes are measurable. What follows isn’t a projection. It’s documented evidence from organizations that chose to act.

eHealth NSW: 144,000 hours of clinical productivity recovered annually in a single health system migration
99.99% system availability in cloud compared to 99.5% on-premises, 43 additional hours of uptime per year, per hospital
In one of the largest cloud migrations ever undertaken by the NHS, Spine’s 36 national services—hosted on more than 360 servers across two UK data centers and containing more than 1.2 petabytes of data— were migrated to the AWS Cloud. The migration was accomplished with zero downtime, ensuring availability and continuity of services for all end users.

“We are convinced that with the help of AWS we can make an essential digital transformation, not only for the hospital but also for the national health system.” – Dr. Josep Maria Campistol, CEO Hospital Clinic Barcelona

Conclusion

The journey from EMR systems to generative AI solution isn’t a replacement story. It’s an augmentation story. EMR systems on AWS serve as the trusted, governed, and increasingly standards-compliant repositories that feed a new generation of clinical AI agents.

Key takeaways

Healthcare organizations that have migrated clinical systems to the cloud are already reporting measurable gains across productivity, availability, cost, and recovery:

The data readiness of your EMR system directly determines the quality and safety of your AI outcomes
AWS provides two complementary paths: extend proprietary EMR capabilities with AWS AI services, or build on open standards (FHIR, openEHR) for maximum interoperability
The commitment AWS has made to FHIR (through AWS HealthLake), openEHR (Diamond Membership), and EHDS compliance makes it the natural platform for European healthcare AI
Amazon Connect Health demonstrates what’s possible when EMR data is properly structured, semantically enriched, and made AI-ready

Why this matters

We can measure this journey in uptime percentages and cost reduction ratios. We can measure it in FHIR compliance scores and interoperability benchmarks. These metrics matter. But the measure that matters most is simpler: does the clinician walking into the room have everything they need to care for the person in front of them? Does the patient feel seen—not as a record number, but as a person with a history, a context, a story?

Every patient is an algorithm of DNA, dreams, and history. A unique life story behind each diagnosis. Cloud-centered EMR is how we begin to honor that, improving patient and clinician experience, advancing population health, reducing inequity, and making the system more sustainable for everyone.

Because every health story matters. Every. Single. One.

Next steps

Explore AWS Healthcare and Life Sciences partner ecosystem, review the AWS HealthLake FHIR capabilities, and consider how your EMR data foundation maps to the agentic AI-ready data stack described in this post:

AWS for Industries