More Than Just A Chat Bot: Exploring FHIR and AI Integration in Healthcare Applications
Artificial intelligence is increasingly being woven into healthcare applications, from patient record systems to operational workflows and clinical decision support tools. But what happens when AI becomes part of the user experience rather than something that sits alongside it?
This series documents a set of observations, lessons and considerations that emerged while building a healthcare AI playground around synthetic patient data. It explores how conversational AI, healthcare interoperability standards and patient records might work together, and examines some of the questions that arise around user experience, trust, accountability and information governance.
The articles are not intended as a blueprint or implementation guide. Instead, they capture some of the practical considerations and discoveries that emerged whilst exploring how AI could be integrated into healthcare applications.
About the Example Application
The articles in this series are based on a healthcare AI playground application that we built to explore ideas, test concepts and demonstrate approaches rather than prove any particular solution.
The application began as a way of working with a range of FHIR capabilities and healthcare interoperability standards, including Structured Data Capture (SDC), GraphDefinition, Terminology Services, StructureMap transforms and PlanDefinition execution. Over time, it evolved into a platform for exploring how AI might be integrated alongside those standards.
Much of the series draws from that second area of work. The focus is not on prescribing a particular architecture or implementation approach, but on sharing some of the observations, lessons and considerations that emerged along the way.
All patient data used within the application is synthetic, generated using Synthea, an open-source synthetic health care data generator. This allows us to show real application output throughout the series without using real patient information.
AI Specific Features
📝 AI Summarisation
✨ Integrated AI Agent
🔗 Citations
📅 Appointment Booking
📂 Document Filing
Example Application
Read the Series
More Than Just A Chat Bot
The opening article explains what motivated the work and introduces the central question explored throughout the series: what happens when AI becomes part of a healthcare application rather than a standalone tool? It outlines the playground we built, why we built it, and some of the themes that emerged along the way.
Read: More Than Just A Chat Bot
Agents Are Users Too
Modern AI assistants do more than answer questions. They may retrieve information, make decisions about how to fulfil a request and even perform actions on behalf of users.
This article explores what that might mean for accountability, auditability and provenance, and considers whether AI agents should be treated as users of a healthcare system in their own right.
Read: Agents Are Users Too
A Chat Interface Is Not a Side Door
A conversational interface may feel very different from a traditional healthcare application, but should it follow different rules?
This article examines how permissions, access controls and patient context apply to AI-powered chat experiences, and why conversational interfaces should not become an alternative path around existing governance controls.
Read: A Chat Interface Is Not a Side Door
Citations Are Not Decoration
Trustworthy AI requires more than convincing answers.
This article explores how citations help users understand where information came from, verify AI-generated responses and navigate back to the underlying patient record. It argues that citations are not simply a user-interface feature, but an important part of transparency and trust.
Read: Citations Are Not Decoration
The FHIR Knowledge Base
Healthcare systems contain vast amounts of structured information, but finding the right information at the right time remains a challenge.
This article explores how standards such as FHIR can provide the foundation for AI-powered retrieval and search experiences, and considers how structured healthcare data might be transformed into a knowledge base that remains anchored to the source record.
Read: The FHIR Knowledge Base
Chunking Is the New Data Modelling
One of the least glamorous topics in AI turns out to be one of the most important.
This article looks at how information is prepared for retrieval, how seemingly small design decisions influence the quality of answers produced by AI systems, and why chunking starts to feel remarkably similar to more traditional data modelling disciplines.
Read: Chunking Is the New Data Modelling
It Takes More Than One Agent
The AI behind a chat interface is rarely a single thing.
This article explores how specialist agents may collaborate to retrieve information, validate responses and perform actions. It examines agent orchestration, workflow design and the practical reality that successful AI integration often depends as much on the surrounding processes as it does on the model itself.
Read: It Takes More Than One Agent
Whether you are developing an AI strategy, modernising healthcare applications, exploring interoperability platforms or looking at ways to introduce AI into clinical and operational systems, come and talk to us. We’d love to hear about the challenges you’re trying to solve and share some of the lessons we learned along the way.
