Who Owns the Patient Relationship Now?

The image I keep returning to is a wall.

On one side: centuries of accumulated medical knowledge, locked inside journals, textbooks, board examinations, clinical guidelines, medical guilds, and a terminology specifically designed to signal that access requires initiation. On the other side: everyone else. Billions of people who, from time to time, would pay for brief, metered passage through the wall to spend fifteen minutes with someone who lived on the other side.

I have been on both sides of that wall. As a physician, I was trained within it. As a product manager who has spent fifteen years watching technology reshape markets, I am watching it crumble in real time.

The experience is something between awe and vertigo.

My argument is this: AI is dismantling the information advantage that organized medicine for centuries, and consumer tech platforms are racing to own the relationship that fills the gap. The question nobody is asking loudly enough is whether that is better for patients, or just different.

The Wall Was Governance

Before I describe the crumbling, I want to say something about the construction. The wall was not built by cynics protecting their income. It was built by people who believed, with good reason, that medicine required protection from charlatans, unqualified practitioners, and incomplete information. The Latin signaled precision. The credentialing was quality control. The professional societies and licensing bodies maintained standards during an era when reliable information was genuinely scarce and the harm from bad medicine was catastrophic.

What I have been calling a wall is, in more precise language, governance. A set of rules about who could access what, who was qualified to provide what, and who was accountable when something went wrong. When a physician caused harm, there was a board to answer to. When a drug failed, there was a regulator. When a diagnosis was wrong, there was a standard of care against which the decision could be measured. It was imperfect governance, captured at times by professional self-interest. But it was governance, with a defined accountability structure underneath it.

For a long time, it served its purpose. The asymmetry between physician and patient was real, and the architecture around that asymmetry was a reasonable response to it.

What changed is not the good intentions of the people inside the wall. What changed is the information asymmetry itself. When a patient can access detailed, referenced reasoning about their symptoms, their medications, and their test results without a physician's permission, the structural premise shifts. LLMs have not yet transformed clinical outcomes at scale, and the evidence on that is still forming. But they have already changed access. For an increasing share of what happens inside the wall, the wall is no longer the only way through.

The governance, however, has not been replaced. That is the part worth watching carefully.

The Biggest Blue Ocean in History

As a product manager, the framework I reached for when I watched this happen was "Blue Ocean Strategy." Kim and Mauborgne describe blue oceans as uncontested market space, places where competition is irrelevant because the playing field has not yet been defined. The collapse of the physician knowledge monopoly created the largest one I have seen in my career.

Ninety percent of US healthcare expenditure is chronic disease management. $5.3 trillion annually. Every person. Lifelong engagement. The most personal data in existence. And for the first time in history, no incumbent with the resources, the incentive structure, or the platform relationships to capture it at consumer scale.

The blue ocean that opened when the wall began to crumble is not a niche opportunity. It is, by most measures, one of the largest untapped engagement markets in the history of technology.

The Race to Turn It Red

According to the AMA's 2026 physician survey, 81% of physicians now use AI in some form, up from 38% in 2023. Most of that use is administrative: documentation, scheduling, summarization. Clinical decision support is further behind. But the directional signal is clear enough: the resistance argument is effectively over.

But what I want to focus on is not the clinical adoption curve. It is what Amazon, Google, Microsoft, OpenAI, and Apple are all doing simultaneously, and why it looks so similar across all of them.

Amazon: AI entry point to One Medical to Amazon Pharmacy. Google: foundation models scoring at physician level on standardized medical benchmark examinations, search embedded in the clinical data layer. Microsoft: ambient AI inside the EHR, passively listening to every clinical encounter. OpenAI: ChatGPT with health record integration, with over 230 million weekly users globally and a growing share asking health-related questions. Anthropic: Claude with strong medical reasoning capabilities, increasingly embedded in health workflows and clinical decision support. Apple: the longitudinal health profile assembled from every device the patient carries.

These are not different bets on different futures. They are the same bet. Every major platform has identified the same opportunity: own the interface layer between the patient and every health decision they make, and capture the longitudinal relationship that makes all downstream data flow to you anyway.

Geoffrey Moore's "Dealing with Darwin" is useful here. The companies that survive Darwinian competitive disruption are not the ones that compete hardest on the existing playing field. They are the ones that understand what type of innovation they are actually executing. Tech giants are not trying to build better hospitals. They are executing category creation, a move that renders the existing competitive landscape irrelevant by defining a new one. Blue Ocean Strategy describes what happens when everyone identifies the same new category simultaneously. The ocean turns red. We are watching it happen.

Why Cooperation Is the Wrong Frame

Here is the argument I want to make directly: anyone who believes the relationship between tech giants and traditional healthcare will resolve into genuine synergy is misreading both the incentive structure and the historical record.

The partnerships are real. The integration announcements are real. None of it is theater.

But read the data clause, not the press release.

The historical pattern for category creation moves against incumbent industries is consistent enough to be predictive. Music cooperated with Spotify on distribution and lost margin capture. Retail cooperated with Amazon's marketplace and lost category after category. Travel cooperated with Google on traffic and lost the booking relationship. In every case the cooperation was real, the value extraction was structural, and by the time the incumbents understood what had happened, the platform owned the customer.

Healthcare will likely follow the same arc, slower because regulation, clinical trust, and the genuine complexity of medical care create real friction that did not exist in music or retail. But the direction is the same. The cooperation buys legitimacy and clinical access while the platform builds the alternative engagement layer. When that layer is rich enough, the cooperation becomes optional. The health system becomes the high-cost backend for the cases the platform cannot handle.

The Second Wall Nobody Is Talking About

The knowledge wall is crumbling. There is a second wall still standing: the data.

Healthcare IT fragmentation is not a design failure. It is a design feature, built deliberately by rational actors protecting rational interests. Epic controls roughly 35 to 40 percent of US hospital EHR market share through a business model built entirely around being the irreplaceable system of record. Every proprietary portal, every incompatible lab system, every imaging report delivered as a PDF: these are fragments of a data wall that serves the same function the knowledge wall did. Own the asset everything else depends on. Make sure it never flows freely to anyone who might use it against you.

The 21st Century Cures Act mandated interoperability. FHIR APIs exist. And the practical experience of moving patient data across systems remains exactly as difficult as before, because compliance and interoperability are different things, and the industry has optimized for the former while structurally resisting the latter.

What the tech giants are building is an attack on this wall from the patient side. Apple Health pulls hospital records directly to the patient's device through FHIR patient access APIs. AI platforms can analyze those records once the patient shares access. Amazon Health AI assembles a consolidated longitudinal profile from every source the patient connects.

But the most powerful mechanism driving this is not regulatory compliance. It is usefulness, delivered through interfaces that feel nothing like a hospital portal.

When an AI platform can answer specific questions about a patient's own medications, flag an interaction with their actual prescription list, explain a lab result in plain language, or connect a symptom to something in their history, patients will share their records willingly. Many patients who would hesitate to hand their medical file to an insurer will grant a consumer AI platform access to their clinical history because the interface is empathic, the response is immediate, and the value is concrete. Claude, ChatGPT's health features, and Amazon Health AI are all converging on this model: not mandating access to patient data, but making the value of sharing it obvious enough that patients volunteer it without being asked twice.

Privacy concerns do not disappear in this transaction. They are simply outweighed, in the patient's calculation, by the experience of being understood by a tool that actually knows their history. That is a more durable attack on the data wall than any regulatory framework, because it operates through patient choice rather than institutional compliance. Regulatory mandates can be slowed, litigated, and minimally satisfied. Demand cannot.

One Medical, the clinical backbone of Amazon's health funnel, does not run on Epic. It runs on a proprietary EHR platform, 1Life. Every patient Amazon captures through its funnel has their longitudinal clinical record in 1Life, not in the hospital ecosystem that feeds the traditional health IT stack. Every new Amazon health customer is clinical data moving permanently out of the fragmented incumbent infrastructure and into a platform Amazon controls. That is not a side effect of the strategy. It is the strategy.

The knowledge wall began eroding through search and telemedicine and open-access publishing, and accelerated sharply with LLMs. The data wall will come down differently: patient by patient, record by record, as platforms make it incrementally easier to move data toward the interface people already trust. The trajectory is the same, the mechanism is slower, and the incumbents have more time to respond. Whether they use it is a different question.

What Good Design Actually Requires

I tested Amazon's Health AI recently, not by design. It surfaced on my Amazon homepage while I was managing a headache. The experience was smooth. The AI was competent. What struck me was not the quality of the clinical reasoning. It was what the product architecture was optimizing for.

Good product design is honest about its incentives. The best consumer health platforms will be the ones that make those incentives aligned with patient outcomes: real data portability, transparent routing logic, clear accountability when the recommendation causes harm, and no steering toward proprietary services when a better option exists.

The difference between a platform that empowers patients and one that captures them is not in the press release. It is in specific product decisions made during design reviews, by product managers asking the right questions or not.

I have watched from both sides of a wall that separated the people who held medical knowledge from the people who needed it. I watched the knowledge asymmetry that justified that wall dissolve faster than anyone in medicine predicted. I am now watching the data asymmetry that replaced it come under the same pressure, from a different direction and with much more money behind it.

The wall is crumbling. What was built in its place, so far, is engagement infrastructure without accountability infrastructure. Platforms without governance. Scale without standards.

The question is not whether the old wall should have stayed standing. It should not. The question is who is building what comes next, and whether they are optimizing for the patient's outcomes or their conversion metrics.

Those are different things. And unlike the wall, the difference will not take centuries to become visible.

This article is the third in a series on the changing architecture of patient care. Previous pieces: "Patients Are Not Waiting for Permission" and "The Parallel Health System."