There is a person we should talk about before we talk about AI. She is a Regional Director at a mid-sized home care agency. She gets in before 7am. She leaves after 6pm. Not because she loves spreadsheets. Because somewhere between the scheduling gaps, the callouts, the certification expirations, and the billing exceptions, there is a patient who needs a caregiver to show up tomorrow morning. That is the job she came to do. The administrative work is what prevents her from doing it.
Everything that follows is about her. About the technology choices being made on her behalf, without her input, that will determine whether she spends the next five years closer to that patient or further away.
Your EMR has an AI strategy. It was decided in a product roadmap meeting you were not invited to. And because switching your EMR costs somewhere between three hundred and twenty million dollars and three years of organizational upheaval, their strategy is now your strategy. Permanently. Until further notice.
This matters more than most operators realize. Here is why. And here is what history has to say about it.
Three Empires. Three Choices. One Lesson.
In the early 1600s, the Dutch Republic was a small, waterlogged nation with no army worth mentioning and no natural resources of any kind. Then it made a decision that changed the world.
When the Spanish Crown began expelling Jews and Protestant Huguenots from across Catholic Europe, Amsterdam said: come here. When skilled weavers fled religious persecution in Antwerp, Amsterdam said: bring your looms. When merchants, scientists, and philosophers needed somewhere to think and trade freely, Amsterdam opened its doors and meant it.
The result was the Dutch Golden Age. Within a generation, Amsterdam became the financial capital of the world. The Dutch East India Company became the most valuable corporation in human history, worth an estimated $8 trillion in today's dollars. Rembrandt painted. Spinoza philosophized. Dutch ships controlled the global spice trade. A nation of two million people ran the world's economy for a century. Open the doors. Welcome the newcomers fully. Thrive beyond all proportion to your size.
Two centuries later, Japan sat at a different crossroads. In 1853, Commodore Perry's Black Ships arrived and made plain what isolation had cost. The Meiji Restoration of 1868 responded with a different answer. Japan didn't simply open its doors and let newcomers run things. The emperor's Charter Oath was precise: "Knowledge shall be sought throughout the world so as to strengthen the foundations of Imperial rule."
Japan hired over 3,000 foreign experts, the oyatoi gaikokujin, and paid them accordingly. In 1874, their salaries consumed 33.7% of the national annual budget. Japan learned everything they knew. Then sent them home. By the mid-1880s, most were gone, replaced by Japanese engineers who had internalized and surpassed them. In four decades, a feudal agricultural nation became an industrial world power. Welcome newcomers temporarily. Extract their knowledge. Build everything yourself. Graduate from foreign help entirely.
And then there is the third path. The one no empire ever intends to take.
In 376 CE, tens of thousands of Goths massed along the Danube. Not invaders. Families, fleeing the Huns. They begged to enter Rome. Rome had a playbook that had worked for centuries: absorb newcomers, fold them into the legions, grow stronger from the exchange. It was the greatest integration machine in history.
But the officials in charge were corrupt. The protocols were ignored. The Goths were not integrated and not turned away. They were stranded in the bureaucratic middle: acknowledged but unserved, promised safety and given starvation. In 378 CE, they annihilated a Roman army of 40,000 at Adrianople. Thirty thousand Romans died in a single afternoon. It was the beginning of the end of the Western Empire. They chose neither path clearly. Managed the middle. Rot from within. Fell.
Epic Is Japan. And That's Correct.
Let's talk about Epic.
Epic Systems commands 42.3% of the U.S. acute care hospital market - its largest share ever, still growing, winning nearly 70% of all new hospital EHR decisions in 2024. It covers 54.9% of acute care inpatient beds in this country. Its Cosmos database holds 16 billion clinical data points. It generated an estimated $5.7 billion in revenue in 2024, up 16% year over year. Patient records for 325 million people live on its servers.
Epic is not a software company. Epic is infrastructure. It is roads, sewage, and the power grid rolled into a single product that hospitals cannot leave even when they hate it.
And when AI newcomers started arriving at Epic's gates - ambient scribes, intelligent agents, scheduling algorithms, care coordination tools - Epic made the Meiji calculation. We have the resources. We have the data. We have the distribution. We will build it ourselves.
This is the Meiji strategy executed with Silicon Valley money and Big Ten institutional arrogance. The foreign experts - Abridge, Nuance, Ambience, a hundred others - have been welcomed at the conference, studied, and are now being methodically replaced by internally built equivalents. Epic's App Showroom has over 1,000 live vendor apps. FHIR API calls surged from 6 billion to 10 billion monthly calls in a single year. They let the newcomers in just long enough to understand what they were building. Now Verona is building it themselves.
I am not criticizing this. With Epic's resources, scale, and data moat, the build strategy is correct. It is arguably the only strategy that is honest at their size. When you hold the records of 325 million Americans, you owe your customers integration, not dependency on a fragmented third-party ecosystem. Epic is doing what Japan did. And it will work.
"Epic can afford to be Japan. The question is whether post-acute and home care EMRs are pretending to be Japan when they're actually late-stage Rome."
Post-Acute and Home Care Has No Epic. This Changes Everything.
Now walk to the other side of the discharge summary. Back to our Regional Director and her 7am shift gap.
In post-acute and home care, there is no Epic. There is not even one company with 40% of the market. The five major players, PointClickCare, Homecare Homebase, MatrixCare, WellSky, and Netsmart, each dominate a different slice. Homecare Homebase leads in Medicare-certified home health by a wide margin, with its president stating in April 2025 that HCHB serves roughly 45% of the Medicare home health market, with $26 billion in claims flowing through the platform. Together, the top five control roughly 45% of total home health software revenue in a market worth approximately $4.5 billion and growing. Compare that to Epic's $5.7 billion generated from a single coherent platform. This is not a monoculture. It is a confederation of regional dialects trying to pass as a language.
PointClickCare does not have a Cosmos, though it claims the largest post-acute dataset in North America and has built a 400-partner marketplace that signals a real openness strategy. MatrixCare does not have a $250 million ambient documentation partnership with Microsoft. Netsmart cannot build 125 generative AI features simultaneously while managing the integration debt of several prior acquisitions on a fraction of Epic's budget. Homecare Homebase processes $26 billion in claims annually and serves 300,000 users, yet has no public AI development program at scale. These are capable companies serving a sector that is harder, more underfunded, and more regulation-laden than acute care. That is simply the truth.
But the build strategy is not available to them. Not at the pace AI is moving. Not at the quality operators need. Not without taking their eyes off the compliance updates, EVV mandates, and PDGM changes already congesting the roadmap.
When a post-acute EMR says "we're building AI," what it typically means is a small feature team adding a predictive readmission score to a product backlog that is already two years deep. That is not building AI. That is taping a satellite dish to a farmhouse and calling it a space program.
Japan spent 33.7% of its national budget to execute the build strategy. It was total commitment with total clarity of purpose. A mid-market post-acute EMR with a modest R&D line trying to replicate what Abridge raised $460 million to build is not Japan. It is the worst version of Rome: managing the middle at the Danube, with everyone pretending the Goths are not already camped on the other side.
Two Paths Worth Taking for EMRs.
Here is the honest frame for what post-acute operators should be asking of their EMR partners right now. Not as a vendor evaluation. As a strategic demand from a customer who cannot afford to wait.
There are two paths that work in this market. Both are honorable. Both create real value for operators. The AI companies arriving at post-acute EMR gates right now would thrive under either one.
Path One: Become the most open EMR in the market. Build real-time, well-documented APIs for scheduling data, caregiver availability, visit verification, and care plan management. Stop charging integration tolls to companies trying to serve your customers. Make it fast and cheap for best-in-class AI to sit inside your platform. Be Amsterdam. PointClickCare's 400-partner marketplace is a genuine step in this direction: operators choose it partly because it makes their whole stack work better, not just the parts PointClickCare built. That is the right instinct. Scale it. Deepen it. Make the API as easy as the UI. Openness is not a concession. It is a strategy.
Path Two: Out-innovate at a pace that earns trust. Make real bets. Not a demo at the user conference. Not a beta for select customers that never reaches general availability. A product your operators use every Tuesday at 6am when they are trying to cover the morning's open visits. Ship it. Learn from it. Ship the next one. The build strategy is available to post-acute EMRs, but only if it is executed with the urgency the moment demands, not added to a roadmap and revisited quarterly.
These two paths can coexist. The EMR that is both open and fast is the one that wins this decade. The worst outcome is neither: the innovation theater, the API that technically exists but practically frustrates, the partnership announcement that never becomes a shipped product.
The Lease You Signed Without Reading It.
Here is what makes this more than a vendor conversation.
EMR switching costs are among the highest of any software category in any industry. Trinity Health is completing an $800 million Epic implementation across its network. Sarasota Memorial is spending $160 million over five years. Inspira Health: $120 million. These are acute care numbers. But the principle does not shrink with the market. Switching an EMR for a 300-location home care network is a three-year transformation that touches billing, clinical workflows, compliance, payroll, and the institutional memory of every coordinator who has been doing it the old way for a decade.
What this means - and this is the point I want you to sit with - is that your EMR's AI strategy is your AI strategy. Whether you like it or not.
There is no version of this where you operate on one platform and build an independent AI strategy on top of it without friction. The data lives in the EMR. The workflows live in the EMR. The compliance reporting lives in the EMR. If your EMR is moving slowly on AI, you are moving slowly on AI. If your EMR treats third-party AI vendors as threats rather than partners, your operators are locked out of tools that could save them millions in turnover costs, fill rates, and overtime spending.
You have signed a long-term lease on someone else's AI ambition. Most operators don't realize this until the lease is five years deep.
The Choice in Front of Operators.
If you're an operator with leverage - a large network, a significant contract renewal, a seat at the user advisory board - you have real power here. Use it.
The first move is to make your EMR partner faster. Demand a named AI roadmap with dates, not categories. Ask which specific AI capabilities will be live in your environment - not "generally available," not "in beta for select customers" - in your environment, in the next eighteen months. Push for contractual commitments on API openness. Organize with other large customers and speak with one voice. User conferences are not listening sessions. They are leverage events. Use them.
The second move is to make your EMR more open. Specifically request published, well-documented APIs for scheduling data, caregiver availability, visit verification, and care plan management. Ask whether the platform will support deep integrations with the AI vendors your organization wants to use - not just the ones the EMR certifies through its own marketplace. Openness is not an optional feature. It is a fiduciary obligation to operators trying to compete.
Anything in the middle is Adrianople.
The EMR that says "we're working on it" while not shipping, that invites AI vendors to the conference but charges extraction tolls for integration access, that puts AI on the roadmap but not on the release calendar - that EMR is choosing the path of the Western Roman officials who starved the Goths at the Danube. Not maliciously. Not strategically. Just bureaucratically. Just slowly. Just without the urgency the moment demands.
And you know what happened to the Western Roman Empire.
"The EMR that chooses neither path - not open enough, not fast enough - is not being cautious. It is choosing the outcome it claims to fear."
The AI newcomers at the gate right now are not here to replace what your teams do. They are here to give your Regional Director her mornings back. To make sure the caregiver shows up tomorrow. To move the paperwork out of the way so the care can get through. That is the only reason any of this matters.
The EMRs that open their doors will look, in ten years, like Amsterdam at its peak: the center of gravity for everything that matters in this industry. The ones that try to build everything themselves without the resources to do it will look like what they are: a feudal castle with a satellite dish on the roof. And the ones that choose neither path, that acknowledge the newcomers without embracing them, that promise speed without delivering it, that sit at the Danube collecting tolls from companies trying to serve their own customers: they are writing the final chapters of an empire that does not yet know it is already falling.
Open the gates. Move fast. The newcomers are already here. And your patients are waiting.
