Physicians spend more time documenting care than delivering it. Epic built the system that made that true, and Epic is the only organization with the scale and the data to fix it.
This is how I would approach that redesign, and why the people currently doing it are solving the wrong problem first.
The Numbers Behind the Most Expensive Usability Failure in History
Start with what the data actually says, because the scale of this problem is not adequately represented by the word “disliked.”
According to the AMA’s 2024 Organizational Biopsy, physicians work an average of 57.8 hours per week. Of those hours, only 27.2 are spent on direct patient care. Thirteen hours go to indirect care activities including order entry, documentation, test result interpretation, and referrals. Another 7.3 hours disappear into administrative tasks. Physicians are spending less than half their working hours on the thing they trained for. The EHR is the primary infrastructure of the hours they are not.
A 2025 national survey by Black Book Research found that 92 percent of nurses believe EHR systems have negatively impacted their job satisfaction. Nearly 40 percent of their shifts are now spent on documentation instead of direct patient care. One-third of nurses experiencing burnout cite EHRs as a major contributor, and 40 percent of those are likely to leave their organization within two years. Documentation and charting ranked as the number-one driver of physician burnout in Tebra’s 2025 Physician Burnout Survey, cited by 26 percent of primary care physicians.
A systematic review of 29 studies across hospital settings found that usability issues and time spent on the EHR were the most significant predictors of clinician stress and burnout, more significant than any other measured variable. A scoping review published in 2025 found that deep navigational hierarchies and nonintuitive menu labels doubled the clicks required to reach basic documentation functions, with wrong-field data entry occurring in 17 percent of observed tasks.
And an implementation study of Epic in Denmark and Finland found that five and three years respectively after go-live, 32 percent of users remained dissatisfied or very dissatisfied with the system, the time required to perform common clinical tasks had increased, and usability problems had been slow to correct.
Epic powers over 250 million patient records and holds between 37 and 39 percent of the US acute care hospital market. It is the operating system of American healthcare. It is also, by the measure of its primary users’ experience, a system that makes the most important work in the world harder to do.
That is the design problem I would start with. Not the features. The philosophy.
What Every UX Boom Taught Us About Systems Built for Administrators, Not Users
Every major UX paradigm shift reveals, in retrospect, who the design was actually optimized for. Epic’s history is instructive here.
The GUI era of the late 1980s and 1990s established the screen-based interface as the primary mechanism for human-computer interaction. When healthcare began its digitization process in the early 2000s, the EHR inherited this paradigm wholesale. The screen, the menu, the form field, the click path: these were the design primitives of a generation of medical software built not around how clinicians think or how care is delivered, but around how computers organized data. The system was designed to capture information for billing, compliance, and record-keeping purposes. The user experience was a secondary consideration at best.
The mobile revolution of the late 2000s and 2010s produced a generation of consumer software that reset expectations for what usable technology felt like. People began arriving at clinical settings having spent their entire digital lives with apps that were fast, intuitive, and designed around their actual behavior. Epic, built on an architecture developed in the 1970s and evolved incrementally for decades, felt the contrast acutely. The gap between what technology felt like in every other context of a clinician’s life and what it felt like inside an Epic Hyperspace session became the defining usability grievance of a generation of healthcare workers.
The conversational UI era of the 2010s began providing early answers. Ambient documentation pilots using voice AI showed that removing the keyboard from the documentation loop produced meaningful reductions in after-hours work, note completion time, and reported burnout. The research was directional before it was comprehensive, and Epic watched it carefully.
Now, at the beginning of the ambient intelligence era, Epic has announced the most significant design evolution in its history. The question is whether the evolution is deep enough.
Why the Ambient Intelligence Era Is Epic’s Redesign Opportunity and Its Existential Risk Simultaneously
Epic is moving. At its 2025 user group meeting, the company announced Art, an ambient AI scribe integrated with Microsoft’s Dragon technology that converts physician-patient conversations into structured clinical documentation in real time. At HIMSS 2026, Epic previewed an agentic AI roadmap with new capabilities across clinical, patient-facing, and operational workflows. Cosmos AI, trained on more than 8 billion patient encounters, is being positioned as a predictive care and research platform. A redesigned clinician and patient interface is scheduled for 2026.
These are real developments and they represent genuine progress. Epic is rolling out ambient documentation directly into the workflow, and 85 percent of its healthcare customers are now live with generative AI across its Art, Emmie, and Penny AI assistant tools. The ambient AI market is now the most successful AI use case in healthcare, with investors committing nearly $1.6 billion in 2025 alone.
But ambient documentation is not a redesign of Epic. It is a new input method layered onto an existing architecture. The documentation burden is addressed. The navigational complexity, the alert fatigue, the workflow fragmentation, the interface designed for data capture rather than clinical judgment: none of these are solved by adding a microphone to the session.
Here is how I would approach the redesign that the documentation AI is making possible but not completing.
The Three Shifts I Would Build the Redesign Around
Shift 01: Redesign for clinical judgment, not data capture
The fundamental design error of every EHR ever built is that the interface was organized around what the system needed to record rather than what the clinician needed to decide. Epic’s navigation structure reflects the data architecture of a medical record. It does not reflect the cognitive architecture of a clinical encounter. A physician seeing a patient with chest pain does not think in tabs. They think in differential diagnoses, probability assessments, decision trees that branch and recombine based on emerging information. The interface should mirror that cognitive structure.
A redesigned Epic organized around clinical reasoning rather than data capture would surface information in the sequence that clinical judgment requires, not in the sequence that database schema prefers. It would bring relevant prior history, recent labs, active medications, and pending results into a single contextual view assembled around the presenting problem rather than requiring the clinician to navigate to each data source separately. Ambient AI can capture what happens during the encounter. The redesign challenge is building an interface that informs the encounter while it is happening, in a form that reduces cognitive load rather than adding to it.
Shift 02: Design intelligent silence into the alert system
Alert fatigue is one of the most documented and least solved problems in healthcare UX. Epic generates thousands of clinical decision support alerts per clinician per year, and the research consistently shows that the override rate for these alerts is between 90 and 96 percent. A system whose guidance is ignored 90 percent of the time has failed as a decision support system. It has not failed as an alert generator. It is performing exactly as designed: maximizing alert generation without regard to whether the alerts are surfaced at the right moment, in the right form, to the right clinician, for the right patient.
In my redesign, alert architecture would be treated as a signal quality problem, not a volume problem. The question for every alert is not whether the clinical condition it references is real. It is whether surfacing this alert to this clinician at this moment in this workflow will change a clinical decision for the better. Alerts that do not meet that bar belong in a review queue, not in the clinical interface. Ambient intelligence and behavioral data can train a system to know when a specific alert type has a meaningful effect on a specific clinician’s decision-making and when it has become background noise. Building that intelligence into the alert architecture is not a feature addition. It is a redesign of the entire approach to clinical decision support.
Shift 03: Build the interface around the patient, not the record
Epic’s current architecture puts the medical record at the center of the interaction. The patient is, in a functional sense, an input to the record rather than the purpose of it. A redesigned Epic would invert this relationship completely.
The patient encounter view would be organized around the patient’s current situation, history, and active care goals as a coherent narrative rather than as a collection of data fields. Relevant information would be prioritized by clinical relevance to the current encounter, surfaced proactively by the system based on what it knows about the presenting problem, and presented in a form that serves comprehension rather than documentation compliance. The documentation layer would exist, but it would be populated by the ambient AI in the background while the clinician’s attention remains on the patient in the room.
This is the redesign that Epic’s ambient AI investment makes possible. The documentation work is removed from the foreground of the clinical encounter. What should replace it in the foreground is not a blank screen waiting for the next click. It is an intelligent, contextual, patient-centered interface that gives the clinician what they need to make better decisions and then gets out of the way.
The Closing That Should Unsettle Everyone Building Health Tech
Here is the honest assessment of where Epic is right now.
The company understands the problem. The investments in ambient documentation, the redesigned interface scheduled for 2026, the Cosmos AI platform, the agentic AI roadmap previewed at HIMSS 2026: these all reflect an organization that has read the burnout research, understood the cost of its current usability deficit, and committed to changing it. The message at Epic’s user group meeting was clear: the question is no longer how to digitize healthcare. The real question is how to redesign it.
That framing is correct. The execution challenge is ensuring that the redesign goes deep enough.
Adding AI to a system whose underlying design philosophy is wrong produces a faster version of the wrong thing. The documentation burden was a symptom. The disease is an interface architecture organized around data capture, administrative compliance, and billing infrastructure rather than around the cognitive needs of clinicians delivering care under pressure to patients whose lives depend on the quality of those decisions.
Epic has the data to build the most contextually intelligent clinical interface in the world. It has 8 billion patient encounters in Cosmos. It has behavioral telemetry from tens of thousands of clinicians. It has the market share to make a redesigned interface the new standard of care in healthcare software. And it has, for the first time in its history, the AI capability to make ambient intelligence, emotional context, and zero-UI the foundational layer of a clinical experience rather than a feature on top of an existing one.
The tools are here. The only question is whether the design philosophy is willing to follow them all the way.
Build for the clinician in the room. Not for the record being created about the room. The difference between those two design briefs is the difference between a product that contributes to burnout and one that relieves it, and that difference is measured in clinician careers, patient outcomes, and the long-term viability of a healthcare workforce that is already at its limit.
Research sources: AMA Organizational Biopsy 2024, American Medical Association; Black Book Research National Nurse EHR Survey 2025; Tebra 2025 Physician Burnout Survey; Sagepub EHR Stress and Burnout Systematic Review 2023; Wiley Journal of Evaluation in Clinical Practice, EHR Usability Scoping Review 2025; ScienceDirect Epic Implementation in Denmark and Finland 2022; Advisory.com Epic UGM 2025 AI announcement report; Fierce Healthcare Epic AI Charting launch February 2026; Healthcare IT Leaders Epic AI roadmap 2025; Fierce Healthcare HIMSS 2026 Epic agentic AI preview; SPsoft Epic EHR AI Trends 2025; Arkenea Epic Implementation Guide 2026.