How Electronic Health Records are Evolving in the AI Era

The role and functionality of Electronic Health Records (EHRs) is evolving at a dizzying pace.

AI is transforming both how patient data is stored and how healthcare staff and patients interact with clinical data.

This technological evolution is completely overhauling the capabilities and sophistication of these systems, steadily rendering them unrecognisable from their original iterations.

This transitory state is exactly why EHRs have become such a hot topic across the healthcare sector. The specifics of their future trajectory remain uncertain, but industry figures with inside knowledge of the technology's development are keen to share where they think it is heading.

Passive documentation to active intelligence

A recurring viewpoint is that EHRs are being fundamentally repositioned – from a passive system of documentation into an active, intelligent platform embedded in the work of care itself.

Michelle Carmen, Global EHR Practice Lead at EY, describes this transition most explicitly, saying EHRs are evolving “from systems of record into intelligent platforms that actively support care delivery, operational decision-making and patient engagement”. 

She notes that the EHR increasingly functions as "the central hub that connects clinical systems, revenue cycle platforms, patient engagement tools, diagnostic technologies and emerging AI capabilities". 

The diversification of EHRs’ uses means organisations are beginning to view it as "the digital foundation for interoperability, advanced analytics, predictive insights and more personalised patient care", rather than simply a place to record and store clinical data.

Tim Lawless, Global Public Health Lead at Publicis Sapient, also believes this shift is underway. He believes that "the EHR has spent decades becoming very good at documenting healthcare”. 

“The next challenge is making it better at enabling healthcare." He argues AI can turn the EHR "from a system clinicians and patients have to navigate into infrastructure that works more intelligently around them".

DataArt’s Head of Healthcare and Life Sciences, Daniel Piekarz, reiterates this point. “2026 is the year AI agents in the EHR stopped listening and started acting,” he says.

Daniel cites Oracle's Clinical AI Agent as kickstarting this shift, while the subsequent release of Epic and eClinicalWorks’ agent-building platforms reframed the market entirely. 

“For thirty years, what an EHR could do was decided by its vendor,” he explains. “Increasingly it will be decided by the organisations using it.”

Interoperability is a precondition of intelligent EHRs

Intelligent EHRs will also be highly interoperable. An inflection point will arrive when “healthcare organisations will not need to purchase separate AI products for every use case because many capabilities will be delivered natively through the EHR ecosystem”, says Michelle.

She also notes that AI is moving "beyond clinical decision support into ambient documentation, clinical summarisation, coding assistance, prior authorisation automation and revenue cycle workflows". 

However, she stresses this only works if data can move freely: organisations are "investing heavily in FHIR-based architectures, shared care records, and enterprise data platforms because AI is only as effective as the data available to it".

Tim remains cautious, warning that layering AI onto disconnected systems is not a fix.

"AI… exposes weaknesses healthcare has tolerated for years… adding AI on top of fragmented data and workflows doesn't solve the underlying problem; it can simply make a fragmented system move faster."

Michelle also points to a parallel force accelerating EHR evolution: rising patient expectations shaped by consumer technology advances. 

"Patients increasingly expect healthcare experiences that mirror the digital experiences they receive from banking, retail and travel," fueling investment in "digital front doors, self-scheduling, patient portals, seamless communications and easier access to medical information," she explains.

The next 12–24 months: from pilots to platforms

A common theme running through executive viewpoints on EHRs’ medium-term future is that this period is the transition from experimentation to operational, governed deployment at scale.

Michelle sees this as the period's defining shift: "AI will move from pilot programmes to enterprise adoption, with most health systems no longer asking whether they should adopt AI."

Instead, she says, "the conversation will shift to focus on where AI delivers measurable value and how it can be governed responsibly." 

She highlights the rapid expansion across ambient documentation, inbox management, revenue cycle optimisation, prior authorisation and predictive staffing, alongside a parallel rise of enterprise data platforms, where "the EHR remains the system of record, but the data platform becomes the system of intelligence".

Tim is focused on AI implementation quality – not merely AI volume. "The next EHR battle won't be about who adds the most AI. It will be about who can actually redesign the workflow around it," he explains.

He says the current momentum around ambient documentation is on proof of concept, as it addresses an obvious source of friction. “Clinicians [are] spending valuable time documenting encounters instead of focusing on patients." 

Tim again points out that this technology adoption only works if organisations resist the temptation to bolt AI onto existing processes:

"Health systems should be careful not to turn AI into another feature layered onto fragmented processes. Automating a broken workflow doesn't transform it; it can institutionalise the problem." 

Daniel believes that the next phase of EHR development will benefit frontline clinical staff, creating outcomes technology providers may not have initially envisaged. “2027 will be the year when that capability reaches the people who actually understand the work,” he says. 

“When a nurse manager or a revenue cycle director can build an agent for the specific bottleneck they live with every day, you get innovation no vendor roadmap would have prioritised, technology that finally helps deliver care instead of merely reporting on it.”

What does the long-term future hold?

The long-term future of EHRs remains clouded.

Although the speakers’ responses spanned consumer behaviour, regulation and the very location of "intelligence" in the healthcare knowledge ecosystem, each tilted toward the same underlying idea: information is becoming something active that finds and supports people, rather than something people search for.

Michelle referenced the results of a recent EY consumer health survey showing that 76% of those surveyed view AI tools and search engines as reliable sources of health information.

The result, she says, is that "patients are arriving at appointments more informed, asking more targeted questions and expecting more personalised and timely answers".

She frames the broader trend as "the convergence of predictive care, connected care and the rise of the AI-informed patient". She predicts EHRs will shift from documenting what happened to helping clinicians "anticipate what is likely to happen and recommend next best actions". 

Michelle’s also echoes Tim’s points on ensuring AI is properly embedded on a structural level: "The organisations that derive the greatest value from AI will not necessarily be the first to adopt it. They will be the organisations that build the strongest foundation of interoperable, trusted, and well-governed data."

Daniel brings the regulatory dimension into focus, noting that the human oversight obligations of the EU AI Act – the world’s first legal framework governing AI – have taken full effect from August 2026, meaning “clinical decision support sits squarely in its high-risk category". 

As a result, he predicts "a genuinely new role to emerge in health IT" dedicated to auditing agent activity, and the “model to shift quickly from human-in-the-loop, where a person approves each action, to human-on-the-loop, where a person supervises a fleet of agents and intervenes on exception”. 

“That transition, not the AI model capability, will determine how fast health systems can safely move,” Daniel argues.

Tim offers the most structurally distinct prediction, explaining that the biggest change may happen outside the EHR itself, in "a layer of intelligence between the underlying health record and the clinicians and patients who need to use it". 

For patients, he envisions authenticated AI experiences that "securely use a patient's history to help them understand information, find the appropriate care and navigate tasks," and eventually "personal AI agents [that] could begin interacting directly with provider systems on a patient's behalf". 

“That doesn’t make the EHR obsolete,” Tim says. “It makes the EHR increasingly the infrastructure rather than the experience. The real shift is from making people search for healthcare information to making the right information and next action find them."

A word on EHR development for long-term conditions

Nicola Notara, CEO of Vindicara, a specialist tool for diagnosing endometriosis, believes EHRs can play an important role in detecting more chronic conditions. 

“For diseases like endometriosis, important clues can be spread across years of visits, symptoms, diagnoses, medications, imaging and referrals. The data may exist in the record, but the pattern is often missed.

“Over the next 12–24 months, I believe interoperability and smarter clinical decision support will become increasingly important. The opportunity is not to replace clinical judgment, but to organise existing health data in a way that helps clinicians recognise risk earlier and know when a patient may need further evaluation or specialist care.”