Feature by InterSystems
Electronic Patient Records (EPRs) are at a defining point of inflection. For the UK, what began as digitisation of paper records is being reshaped by three converging forces:
- Shifting models of care
- An AI wave
- A policy pivot from acquisition to outcome
The question is no longer whether EPRs will change, but how fundamentally do they need to be reimagined.
The Evolution of EPRs
The adoption and use of EPRs in the UK has significantly evolved from the 1990s through to current day. The transition and adoption of solutions that have moved organisations from the digital filing cabinet through to current day richer clinical platforms that provide a wide range of decision support, service management and planning capabilities, real-time information flows, and service performance insights.
This has reflected a journey across five decades that has seen a transition from paper to departmental solution, through the unified EPR platforms. Interoperability has been a consistent challenge and remains a high priority, and challenge, in enabling the reliable flow of structured and actionable information across healthcare services and systems.
Figure 1 – Five Decades of EPR Evolution
The evolution of the EPR has, to date, been steady and incremental. But taking a step back to reflect upon the last five decades, it feels like we have collectively been building towards this point of inflection.
The Model of Care is Changing
The UK model of care is changing, and EPRs must keep pace and provide an enabling platform upon which the practicalities and logistics of integrated and interprofessional models of care can be delivered. Whilst pertinent for the UK, the changing nature of care is not unique and can be seen in other regions across the EU (e.g. Republic of Ireland with Sláintecare) and further afield. For the UK, the key drivers are
- The NHS 10 Year Plan: care closer to home, integrated care, preventio
- Integrated Care Systems (ICSs): the need for shared service pathways across traditional boundarie
- Person Centric Care: the patient as a participant and not a passive subject of a record
For EPRs, these drivers present critical tensions and challenges:
- Most EPRs were designed for episodic care (predominantly acute) and not longitudinal population health models of car
- EPRs are typically structurally architected to reflect hospital-centric care delivery which increasingly, will prove to be not fit for purpose
The care environment in which EPRs operate is fundamentally and paradoxically changing. The response to the current changing demands therefore requires more than just an incremental tech upgrade.
The AI Wave
The NHS should approach AI not as a single technology decision but as a systemic shift comparable to the move from paper to digital. The difference this time: the pace is compressed, the expectations are higher, and the gap between early adopters and the rest will widen at a pace not previously observed.
A critical question isn’t whether AI will reshape EPRs — it’s whether organisations are positioned to absorb it meaningfully, or whether they’ll more simply layer it on top of systems and processes that weren’t designed to unlock its full potential.
Generative AI has dominated headlines over recent years, but the NHS-relevant applications are more specific and more grounded than the media narrative suggests. Three categories of AI are currently landing with genuine traction in the NHS:
- Ambient clinical intelligence: Ambient voice technology (AVT) that listens during patient encounters and generates structured clinical documentation is moving from pilot to operational deployment. The appeal is clear: reduced screen time, less administrative burden, more clinician-patient eye contact. NHSE has recognised the demand is running ahead of governance and has issued warnings against non-compliant AVT adoption and has launched a self-certified registry for suppliers to evidence compliance (weblink).
- Predictive analytics and early warning: these solutions are typically less visible but potentially more transformative. AI models, including the use of Agentic AI, that proactively flag deteriorating patients, early predictions of sepsis risk, or identify unplanned readmission likelihood are moving from research settings into operational pathways. The challenge here is integration: these tools are only as useful as the clinical pathways they feed into, and most EPRs weren’t designed to surface AI-generated predictions at the point of decision.
- Administrative and operational AI: solutions focusing upon increasing data quality, information processing, and clinical communications whilst reducing the associated resource overhead include clinical coding support, referral triage, discharge summary generation, and theatre scheduling optimisation. Whilst such solutions are less glamorous from a media perspective, these solutions are where the immediate and significant productivity gains sit and where the NHS policy pivot toward productivity and benefits realisation creates a strong business case.
Projecting these three categories of AI onto the NHS 10 Year Plan priorities, there are signposts for where AI can accelerate the delivery of priority service improvements.
- Workforce relief: The NHS workforce crisis isn’t just about headcount, it relates to how clinicians invest their time and how that translates for the delivery of timely, safe, and appropriate care. If, for example, ambient AI can reclaim even 30 minutes per clinician per day from documentation, the compound effect across the system is significant.
- Clinical insight at the point of care: Moving from retrospective data (what happened) to prospective intelligence (what’s likely to happen) represents a fundamental shift that offers new preventative opportunities for care interventions. AI provides the opportunity for EPRs to become a proactive decision support partner to offer synthesis of comprehensive patient records that reflect clinical context and clinician persona perspectives of pertinent information (e.g., highly tailored and dynamic summaries for a surgeon could be different to that of what is required for a community nurse), and the proactive identification of patients at risk of serious complications and deterioration.
- Population health and prevention: At ICS level, AI-driven analytics across linked datasets could enable proactive identification of at-risk populations, targeted interventions, and resource allocation based on predicted demand rather than generic patterns. This maps directly to the care-closer-to-home and prevention agendas.
The Risks and Realities
Associated with the undeniable opportunities offered by AI, there comes a series of critical considerations and challenges that must be tackled head-on for AI to deliver its potential for healthcare and the next generation of EPRs.
- Data quality is fundamental: AI is only as good as the data it consumes. NHS data remains patchy. Coded data varies in quality, free-text records are inconsistent, and interoperability gaps mean AI tools may operate with incomplete pictures. The risk of “garbage in, confident garbage out” is a clear and present risk.
- Bias and equity: Models trained on historically biased datasets will replicate those biases at scale. If AI-driven risk scores systematically underweight certain populations, the consequences in a universal healthcare system are serious. This isn’t theoretical but rather an understood and evidenced risk, for example as described by the recent BMJ article Artificial intelligence and global health equity (weblink).
- Governance and accountability: When an AI-generated recommendation contributes to a clinical decision that goes wrong, who is accountable? The clinician, the trust, the vendor, the algorithm? The current answer is ambiguous, and that ambiguity will slow adoption among risk-averse organisations.
- Workforce readiness: Digital literacy varies across the NHS workforce. Deploying AI tools without investing in training, change management, and clinical engagement risks creating a two-tier system where some clinicians leverage AI effectively and others distrust or ignore it.
- The pace problem: Trusts are being asked to adopt AI while many are still optimising their basic EPR implementations. There’s a risk of running before you can walk by layering AI onto systems that don’t as yet have the quality data foundations to effectively support it.
The Emerging Regulatory Landscape
The regulatory picture is evolving rapidly with regions such as the EU setting the pace. The UK’s framework is taking shape and represents the creation of new regulatory instruments and significant revisions to existing policy and regulations. The tension relates to the need for robust regulatory standards and governance to ensure patient safety and data protection whilst being sufficiently agile to not stifle innovation and adoption. To this end, the MHRA is developing the regulatory framework collaboratively with industry and NHS services rather than a pure top-down enforcement path. Whilst this work continues, software suppliers and NHS services are making investment commitments against a regulatory backdrop that may not conclude until later in 2026.
- National Commission on the Regulation of AI in Healthcare: launched by the MHRA in September 2025, is developing a new regulatory framework to be published by mid-2026. Chaired by Professor Alastair Denniston with England’s Patient Safety Commissioner as deputy, this is the most significant regulatory initiative in NHS AI to date.
- The Commission’s call for evidence (December 2025 – February 2026): addressed three themes – modernising rules for AI in healthcare, keeping patients safe as AI evolves, and clarifying responsibility between regulators, companies, healthcare organisations, and individuals.
- Digital Technology Assessment Criteria (DTAC): is being supplemented rather than replaced. A refreshed AI Buyer’s Guide is in development by DHSC and NHSE.
- Framework procurement process 2026–2027: will allow NHS organisations to adopt innovative technologies including ambient AI through a more streamlined route.
The Policy Pivot: from acquisition to productivity
As recently announced by NHSE, the era of ‘buying EPRs’ is giving way to ‘what are they delivering?’ with an emphasis being placed upon productivity gains and broader benefits realisation as the actual measures of success as opposed to the adherence and achievement of go-live dates.
Figure 2 offers seven different perspectives for benefits realisation, all of which are essential if the true value and impact of an EPR is to be understood and meaningful in the context of continuous improvement. All too often benefits are defined and measured at an organisational or executive level whilst overlooking the those relating to the frontline staff, which can inherently result in significant change management and adoption issues during a deployment. Each level defines “benefit” differently but it requires connecting all seven to form a coherent story.
Figure 2 – Seven Perspectives of Benefits Realisation
The pivot in NHSE policy will increasingly place greater demand upon NHS Trusts to demonstrate and evidence value, and not just achieve implementation schedules. This demand, for some Trusts, will expose both skills and experience gaps in key disciplines relating to change, transformation, and evidence-based benefits analysis. For suppliers, it will further move the conversation with care services away from functions and features to outcomes and optimisation. From a procurement perspective, this would also be a welcome departure, to some degree, away from requirements focussing on the ‘what’ (thousands of disparate functional definitions) to the ‘how’ to emphasise the intended outcome and benefits that the service is targeting. Any productivity gaps will also become glaringly apparent for deployed systems that are not underpinned by an on-going programme of optimisation and continuous improvement.
The emerging policy context pivot will significantly re-shape investment, procurement, and vendor strategy for the incumbent EPR estate and the progression to the next stage of AI enabled platforms.
Two Models of AI-Enabled EPRs : The Strategic Choice
There are two headline conceptual models for an AI-Enabled EPR: Bolted-On and Integrated. Both have strengths and weaknesses and longer-term ability to establish AI enabled capabilities as an embedded facility of healthcare pathways and the delivery of care.
- Bolted-On AI – best of breed AI on a platform EPR
- The current dominant model with AI tools layered onto existing EPR platforms
- Examples include third-party ambient AI, bolt-on analytics dashboards, and API-connected copilots
- Advantages: speed to deploy, vendor choice, lower switching cost
- Limitations: integration friction, data fragmentation, user experience inconsistency, longer-term potential
Figure 3 – EPR with Bolted-On AI : AVT Example
- AI Integrated into the DNA – a platform rethink
- The emerging alternative: EPRs re-engineered with AI at their core DNA
- What this looks like: intelligence embedded in workflow, not added alongside them, and AI enabled actions to, for example, generate actionable orders and tasks from AVT recordings
- Examples of vendors moving in this direction include Oracle Health and InterSystems
- Advantages: seamless UX, deeper contextual intelligence, compound learning, positioned for the longer-term on-going evolution and adoption of AI into healthcare
- Considerations: platform change, scale of change and investment
Figure 4 – EPR with Integral AI Capabilities: AVT Example
The choice isn’t necessarily binary, and organisations may navigate a hybrid path over time to reflect platform capabilities, policy priorities, and service demand. There are also parallels from other sectors (e.g., banking and aviation) where embedded AI is very much redefining their traditional digital platforms.
FROM TOOL TO ASSISTANT — The Future EPR
The conceptual shift of an EPR as a passive repository of documentation to a partner supporting the process of care has been gradually shifting over recent years but will be radically accelerated by EPRs that have AI seamlessly embedded across their platform. We are witnessing the early stages of the emergence of a true EPR ‘AI-enabled assistant’ that will provide a range of core capabilities as set out by figure four.
Figure 5 – Five Key Facets of the EPR ‘AI-enabled Assistant’
The emergence of the AI-enabled platform will present challenges and key considerations that need to be evaluated and resolved in tandem with the introduction of the new technologies. The workforce dimension includes considerations for changes to roles and responsibilities, and digital literacy. For patients, providing clarity as to transparency for the scope of use of personal information, increased shared-decision making, and the greater involvement of care planning through joined-up AI-enabled digital platforms. Underpinning all of these are the foundational needs relating to data quality, governance, trust, and investment.
In Summary…
Whilst AI remains very much in the grips of the hype-cycle, for applications in health care and specifically EPRs, there is sufficient experience and evidence to be confident that ‘The future is bright, and the future is certainly AI-enabled!’, to almost coin a phrase. However, this opportunity demands a very different approach to that what we have witnessed over the last five decades and the evolution of EPRs. An incremental approach is unlikely to deliver the true potential for both EPR supplier and frontline adoption perspectives and may risk the opportunity being perceived as a nothing more than an entertaining technological development. The role of AI therefore sits at the executives tables from national through to regional ICS and individual trusts for strategic planning and enablement with suppliers taking and providing guidance and vision.
The EPR of 2030 will look nothing like the EPR of 2025 and the decisions being made now will determine whether the NHS leads or follows.
