Five scenarios for healthcare delivery, the probabilities behind each, and the indicators that will tell you which future is arriving
Bottom-line first: Healthcare delivery is not going to be replaced by AI. We are asked that question from time to time, or whether every hospital eventually becomes a “server room with an emergency department attached.” That is the wrong fear, and health systems that spend the next decade debating it will miss what actually happens. The more likely outcome is harder to plan for.
WE PREDICT that healthcare in 2040 will be defined by an unbundling that has never happened in the history of medicine: the separation of diagnosis from care. For as long as medicine has existed, knowing what is wrong and doing something about it were fused in the same person, the same visit, the same building. You went to the doctor to learn your condition and to be treated for it. By 2040, diagnosis will be cheap, continuous, and placeless; it will be generated by models and sensors rather than appointments, while care remains physical, local, human, and accountable. The organizations that understand which half of that bundle they actually own will be more valuable than they are today. The organizations still selling the bundle will discover, too late, that half of it has already left the building.
That conclusion comes from looking at what healthcare delivery actually sells, function by function, and asking which functions AI absorbs and which it structurally cannot. It also comes from watching industries that sit eighteen to thirty-six months ahead of healthcare on the AI adoption curve; BigLaw, wealth management, and audit have already run this experiment, and the pattern is consistent: the analytical layer commoditizes first, the work consolidates around a smaller senior judgment layer, and the institutions that treated their analytical edge as the moat are the ones that get repriced.
What Healthcare Actually Sells
Strip the industry down, and it’s clear healthcare delivery sells six things:
- diagnosis (turning symptoms, images, and data into an answer about what is wrong)
- triage and access (deciding who needs what, how urgently, and getting them to it)
- care coordination (moving one patient through a dozen specialists, systems, and handoffs)
- procedural and physical care (surgery, nursing, rehabilitation, i.e. hands on bodies)
- presence (the human being in the room when the news is bad)
- accountability (a named, licensed, insured professional standing behind a life-or-death decision)
AI absorbs the first three almost completely by 2040, and diagnosis is further along today than the rest. Models already match or exceed specialist performance in radiology, dermatology, and pathology reads; continuous sensor data increasingly detects conditions before symptoms present; and the marginal cost of an additional diagnostic inference is approaching zero. Triage and coordination are pure automation targets, and they are where most health systems bleed both margin and patient goodwill today.
The last three resist automation for structural rather than technical reasons. Surgery and nursing require hands. Grief requires presence. And a family authorizing a high-risk intervention needs a human whose license and reputation are collateral – a name on the chart, not a model version number. That need does not disappear when the human stops holding the diagnostic edge; it becomes the entire job. The 2040 physician is a judgment-and-accountability layer, not a diagnostic layer, and the economics of a judgment layer favor fewer, more senior clinicians supervising far more care.
This is the unbundling in one sentence: the half of medicine that made healthcare intelligent is commoditizing, and the half that makes it human is becoming the entire defensible business. Most health systems are investing as if the opposite were true.
The Payer Question
Every few years someone predicts that healthcare data will finally become “liquid”, that interoperability mandates will produce the transparent, portable patient record the industry has promised since the HITECH & CARES Acts. It has never fully happened, and the reason is instructive: the entities holding the data have had every incentive to hoard it. Records were the moat; switching costs were the strategy.
But the more important structural question in healthcare is not who holds the data. It is who pays for the care, because the entity that pays is now, increasingly, the entity that delivers. The largest employer of physicians in the United States is not a hospital system; it is a payer. Vertically integrated payer-providers already own the insurance relationship, a growing share of the clinical workforce, the pharmacy layer, and the analytics infrastructure sitting underneath all of it. When diagnosis unbundles from care, the payers are positioned to capture the diagnostic layer at national scale: they have the data, the actuarial incentive, and the balance sheets to deploy AI diagnosis as a covered benefit long before a regional health system can.
The logic protecting the current provider-centric status quo is weakening for the same reason the information moat weakened in brokerage: AI inference over messy, semi-public data ends hoarding as a strategy. When the diagnostic layer consolidates, it will not consolidate around hospitals. It will consolidate around whoever owns the patient’s data exhaust and the payment relationship, and the institutions it displaces first will be the ones whose value proposition was diagnostic excellence rather than care.
Five Futures, with Probabilities
Scenario 1: The Augmented Clinician (~30%)
The base case, and the winnable one. Healthcare delivery survives as an industry, but the clinician-hours required per unit of care fall dramatically. The traditional apprenticeship pipeline thins; a smaller senior clinical layer operates on top of AI platforms that handle diagnosis, documentation, triage, and coordination. Physician productivity per license multiplies, mid-level scope expands under AI supervision structures, and the clinical shortage that dominates today’s headlines resolves not through more clinicians but through less clinician-dependence.
The dividing line in this scenario is not clinical talent, capital, or brand. It is whether an organization’s clinical and patient-relationship intelligence lives in institutional systems or in individual clinicians’ heads and inboxes. Independent physician groups and regional systems in the second category dissolve or sell cheap when their senior people retire; organizations in the first category absorb their patient panels. This scenario requires no platform monopoly and no regulatory breakthrough; it is simply every organization optimizing independently, which is why it carries the highest probability.
Scenario 2: Payer-Platform Capture (~25%)
The diagnostic layer arrives, but owned by vertically integrated payers and national platforms rather than providers. AI diagnosis becomes a covered benefit delivered through the payer’s own front door; providers become licensed care operators executing treatment plans generated on infrastructure they do not control, paid on rates set by the entity that owns both the diagnosis and the checkbook. Commodity care lines (routine imaging interpretation, chronic disease management, low-acuity primary care) migrate onto payer platforms first. Health systems persist as procedural and acute-care operators, but the intermediary economics (the margin that came from owning the patient relationship end to end) migrate to the platform. This scenario is not speculative; it is the current trajectory of the largest payer-providers, extended fifteen years.
Scenario 3: Care Comes Home (~18%)
Displacement arrives not from payers but from the patient’s own pocket and living room. Continuous monitoring through wearables and ambient sensors, AI primary care as a consumer subscription, at-home diagnostics, and hospital-at-home acute programs move the default site of care out of institutional buildings entirely. The hospital retreats to what genuinely requires it: surgery, trauma, intensive care. Health systems built on outpatient volume and diagnostic imaging revenue discover that their highest-margin service lines were precisely the ones that didn’t need a building. The tell for this scenario is consumer AI health subscriptions crossing from wellness novelty to first-line care relationship at scale.
Scenario 4: Full Agentic Medicine (~7%)
Autonomous AI systems diagnose, prescribe, and manage treatment with humans reduced to procedural execution and final sign-off. Technically plausible by 2040 for narrow domains, (it is arguably already true in corners of radiology), but as a system-wide outcome it requires patients and courts to accept algorithmic accountability for life-and-death decisions, malpractice frameworks rebuilt around model liability, and state licensure boards ceding scope to software. That last requirement is the binding constraint; liability does not automate, and in healthcare the liability is measured in lives rather than basis points. Meaningful probability in commodity diagnostic segments, near zero for acute and end-of-life care.
Scenario 5: Regulatory Freeze (~20%)
Efficiency gains everywhere, structural change nowhere. Reimbursement codes that pay for visits rather than outcomes, state-by-state licensure, liability exposure, and the political untouchability of hospitals as regional employers slow every transformation, and healthcare in 2040 looks like healthcare today with dramatically better documentation tools and a somewhat thinner administrative layer. History gives this outcome more credit in healthcare than in any other industry; this is the sector that absorbed the entire EHR revolution, driving hundreds of billions of dollars of mandated technology, without changing its structure at all. But muddle-through is a probability, not a plan, and it is the only scenario in which doing nothing works.
The Land Is Being Claimed Now
Read the scenarios together and one pattern dominates. In four of the five futures, the diagnostic half of medicine consolidates onto platforms that mid-market health systems and physician groups will rent rather than own; the only disagreement between scenarios is whose platform. And in every one of those four, the durable margin accrues to organizations that own the care half: the patient relationship, the procedural excellence, the local trust, the last mile of medicine that cannot be centralized.
That inverts the strategy most healthcare organizations are currently executing. Boards are approving investments to compete on clinical intelligence (the commoditizing half) while underinvesting in the care experience and relationship infrastructure that will still command margin in 2040. The organizations that survive the unbundling share a common architecture: patient-relationship intelligence that belongs to the institution rather than to individual clinicians, a deliberately senior judgment layer supervising AI-scaled care, and data discipline across their operating systems that most health systems today, running dozens of disconnected clinical and administrative platforms, cannot honestly claim. There is a path to being more valuable in 2040 than today. But this is a land grab with a closing window, and the amount of claimable land is far smaller than the current population of solvent institutions.
What to Watch
Scenario probabilities are only useful if you can tell which future is arriving. Five leading indicators are worth tracking:
- The first CMS reimbursement code for autonomous AI diagnosis. The moment Medicare pays for a diagnostic decision no human made, the unbundling has a business model, and every scenario accelerates.
- Payer-employed share of physicians. Watch the percentage of practicing physicians employed by payer-owned entities; this is the leading edge of the platform-capture scenario, and it is already climbing.
- Malpractice products for AI decisions. When carriers begin writing coverage for algorithmic clinical decisions at scale, the accountability constraint (the binding limit on the agentic scenario) has started to dissolve.
- Hospital-at-home share of acute days. When a meaningful share of acute care is delivered in living rooms, the building-centric health system model has ended, whether or not anyone announces it.
- Consumer AI as first medical contact. When surveys show a majority of patients consulting an AI before any human clinician, and acting on it, the diagnostic relationship has already changed owners.
The Strategic Question
The question for a health system or physician-group leadership team in 2026 is not whether AI will change medicine; that debate is finished everywhere except the industry’s own conferences. The question is which half of the unbundling your organization is building toward – whether your current investments are securing the care relationship you can defend, or merely renting a diagnostic edge that is depreciating toward free. Those are different projects with different architectures, and the organizations that conflate them will discover the difference at the worst possible moment.
Innovation Vista works with healthcare organizations to answer exactly that question: translating what industries eighteen to thirty-six months ahead on the adoption curve have already learned, and turning it into a positioning strategy for the unbundling ahead. If you want to pressure-test which side of 2040 your organization is building toward, that conversation is where we start.
