What the EMR Application Tier Actually Becomes – The Bifurcation, Three Honest Caveats, and Three Questions for Health System CIOs

Post 4 of 4 in The Narrowing EMR series.

KEY TAKEAWAYS

• Epic added 176 net acute care hospitals in 2024 – its largest-ever annual net gain – expanding market share to 42.3%. Oracle Health lost 74 hospitals over the same period. Epic won nearly 70% of new hospital contracts in 2024.
• The argument of this series is late-decade, not near-term. It depends on three conditions maturing between 2028 and 2032: democratized agent-building tooling, AI-assisted semantic-layer extraction at scale, and mature healthcare-native open data platforms.
• The bifurcation: Epic remains the clinical system of record (documentation, orders, MAR, certified quality reporting). What moves elsewhere is analytics, AI agents, population health, clinical research, value-based care risk modeling, and revenue cycle optimization.
• The three questions health system CIOs and CMIOs should be asking in 2026: (1) Where does your data live and who governs it? (2) Who decides what AI agents run against your clinical data? (3) What is your 2030 semantic layer strategy?
• If those proofs of concept for AI-assisted semantic-layer extraction land at Mayo, Kaiser, or Intermountain in the next two to four years, the middle of the market is expected to follow. The question is whether organizations are positioned to move or watching from the bench.
• The decisions you make about what sits around the EHR matter more than the decisions you make about the EHR itself. That has not been true for a generation. In the author’s view, it is now.

Posts 1 through 3 made the analytical case for why the EMR application tier narrows over the late decade and where the strategic value moves. This post is the prescriptive close: what actually happens to Epic, what the bifurcation looks like, and three questions a CIO or CMIO should be working on right now.

Three Honest Caveats Before Any Prescription

Epic is winning the current buyer decision, and the data is not close. Fierce Healthcare, citing KLAS, reported that Epic added 176 net acute care hospitals in 2024, the company’s largest-ever annual net gain, expanding its market share to 42.3%. Oracle Health lost 74 hospitals over the same period, and Meditech lost 57. Epic won nearly 70% of new hospital contracts in 2024. On a 2026 procurement timeline, Epic is often the right call for execution reasons. Nothing in this series argues otherwise. The argument is late-decade, not near-term.

What I’ve laid out depends on three conditions maturing between roughly 2028 and 2032: democratized agent-building tooling, AI-assisted semantic-layer extraction at scale, and mature healthcare-native open data platforms. Each of those is plausible. None of them is guaranteed. If two of the three stall, the timeline slips. If all three accelerate, it pulls forward. A 2026 buyer cannot reasonably bet on any of these landing on a procurement timeline. A 2028 or 2030 buyer can.

Epic remains in the stack, and regulation is part of why. Federal certification requirements, the installed-base legacy of the Promoting Interoperability era, prior-authorization API requirements, and the operational gravity of clinical workflow integration all preserve a floor under the role of the certified EHR. What is changing is the role Epic plays in the stack. The question is whether Epic sits at the center of the hospital’s strategic IT decisions or to the side as the clinical system of record. Those are very different roles. Epic’s posture is not static. The HIMSS 2026 Agent Factory and Curiosity announcements meaningfully expanded what hospitals can build inside the Epic environment. The argument of this series is not that Epic stands still. It is that, based on what is publicly documented today, even Epic’s expansion is happening within the vendor-anchored pole I described in Post 2, not toward the institution-anchored one.

The Bifurcation: What the EMR Application Tier Actually Becomes

Over the next five to ten years, here is what I think actually plays out.

Epic remains the clinical system of record. Documentation, orders, MAR, certified quality reporting, encounter management, and most of the day-to-day clinician workflow continue to run on Epic. Regulation preserves that floor. Operational gravity reinforces it. None of that changes meaningfully on a 2030 timeline.

What moves elsewhere is everything that sits above and around the system of record. The analytics plane. The AI agent plane. Longitudinal population health. Clinical research infrastructure. Value-based care risk modeling. Revenue cycle optimization. Population-scale decision support. That work increasingly runs on an organization-controlled data and AI layer, whether built on enterprise platforms like Snowflake or Databricks, healthcare-specific services like AWS HealthLake or Azure Health Data Services, Oracle’s healthcare data stack, or a healthcare-native successor, with customer-built or customer-governed agents operating against it.

Regulation preserves Epic’s floor. Strategic direction moves it off the ceiling. That is what I mean when I say the EMR application tier narrows. It does not stay where it has been for the last fifteen years. The important shift is that Epic increasingly becomes one critical system in that architecture rather than the unquestioned center of it.

Three Questions Every Health System CIO and CMIO Should Be Asking in 2026

If this argument is right, the work for health system leadership in 2026 is not a different EHR procurement decision. Most organizations aren’t in a procurement window anyway. It’s three questions the dominant strategy of the last decade didn’t require anyone to ask.

Where does your data live, and who governs it? If the honest answer is that most enterprise-critical data decisions still begin inside Epic’s operational data model, that may be an accurate description of 2026 and a poor description of 2030. The work to start now is mapping what an organization-controlled data platform would hold, which decisions that platform would enable that the current architecture does not, and which governance committee actually owns that platform when it exists. The hard work is organizational, not technical. The technical patterns are already proven in adjacent industries.

Who decides what AI agents run against your clinical data? If the honest answer is that Epic heavily shapes the approved ecosystem through programs like Partners and Pals, Connection Hub, and its own first-party tooling, and that your organization does not have a clear answer for the agents your own staff will build, that is worth noticing. It does not mean Epic is doing anything wrong. It means your organization is delegating an increasingly load-bearing strategic decision to a vendor. That may be the right call today. The question worth asking is whether you want to keep delegating it as the agent-building floor drops over the next five years.

What is your 2030 semantic layer strategy? Epic’s clinical logic, decision-support rules, and accumulated configuration are real assets. Reproducing them on a different substrate is not realistic on a 2026 timeline. Over a five-to-ten year window, AI-assisted extraction of equivalent semantic layers becomes plausible, in my view. If I were a CIO planning for that window, I would be watching organizations like the Mayo Clinic Platform, Kaiser Permanente, and Intermountain for early proof-of-concept work in this direction. If credible proofs land, my expectation is that the middle of the market follows quickly. The question is whether you are positioned to move when that happens or watching from the bench.

None of these three questions has a clean 2026 answer. That is the point. The organizations that start asking them in 2026 will have credible answers in 2030. The organizations that don’t will be reacting to the answers other people built.

Where Abundant Fits: Healthcare IT Consulting for Governance and Architecture

At Abundant, we work across the EMR landscape, with deep expertise in Epic, Oracle Health, and the data, AI, and governance layers that sit around certified EHR environments. Most of our engagements run through Epic environments, and we expect that to remain true for the foreseeable future. What this series argues is that the questions we are increasingly being asked in client engagements sit at a layer above the Epic-vs-Oracle decision. Where does the data live? Who governs the agents? What does the 2030 architecture look like? These are governance, architecture, and organizational-capability questions that exist independently of which EMR sits underneath them. That is what we’re built for.

If you are a CIO, CMIO, or CTO wrestling with those questions and the answer cannot be “we will figure it out when the time comes,” it is worth a conversation. Firms like ours are most useful when the question is governance and architecture. We sit alongside whatever Epic optimization or infrastructure execution work the client already has underway, focused on the strategic layer rather than competing with it.

What to take away

The safe path inverted because the environment changed. Reimbursement, care delivery, regulatory interfaces, and AI capability are all in motion simultaneously, and the playbook that won the last twenty years is not optimized for any of them.

Governance is the binding question because technology is no longer the binding question. When agent construction was hard, picking the vendor with the best agents was a reasonable strategy. When agent construction gets easy, the question becomes who governs the agents your own staff will build. That question lives at the data layer, not the application layer.

Healthcare is not exempt from the pattern that played out across every other enterprise data category over the last fifteen years. The 2024-2026 adoption data tells a different story than the legacy “healthcare is slow” narrative does. The pattern is late, not absent.

The EMR application tier narrows. Its role becomes more focused, not less essential. Regulation preserves the floor. Strategic direction moves it off the ceiling. And the health systems that recognize this in 2026 will be positioned for 2030 better than the ones that assume the environment will hold steady.

The argument of this series comes down to a single line. The decisions you make about what sits around the EHR matter more than the decisions you make about the EHR itself. That has not been true for a generation. In my view, it is now.