AI, Humanity, and the Future of Post Acute Care: Where Models Handle Work—So Humans Can Focus on Meaning

Treating AI as an heir to humanity confuses our tools with the people they serve. 

Ambitious, public claims that “superintelligence” could arrive by 2028 produce concern about job loss or at least reconfiguration. Investments match the ambition, with estimates that training a single frontier model can consume tens of gigawatt hours of electricity—enough to power a major city for days. But we see our field missing a key point: increasing scale and energy use can make systems more capable, but without context and integration, do not improve outcomes in real applications. And they do not make models behave more like humans. 

Models, properly deployed, enhance our humanity by taking on the dense, repetitive, computational work that can support use of good judgment, listening, and care. Documentation, order entry, and inbox management are essential for safety, compliance, and payment, but they consume cognitive bandwidth and time. Using technology to deliver these creates space for clinicians and caregivers to do the work that only people can do. Empathy, ethics, and love are not products of scale; they are human capacities. AI can create the conditions for those capacities to show up more often—by giving teams more time, more context, and fewer avoidable errors. 

AI can create the conditions for [human] capacities to show up more often—by giving teams more time, more context, and fewer avoidable errors. 

Consider how we think, versus how we code. A modern model evaluates multiplexed possibilities, assigning weights based on patterns learned from vast datasets, and returns a prediction posed as a solution. The human brain does almost the opposite: it prunes options early, guided by electrical and chemical activity from our most important experiences and governed by our deepest values and desires. Work on the Somatic Marker Hypothesis shows that this “emotional presorting” happens at the outset—from patients with ventromedial prefrontal damage who can reason abstractly but struggle to choose effectively, to Iowa Gambling Task studies where physiological responses flag bad decks long before people can explain why. In this view, emotion is not noise but an efficiency mechanism: feelings narrow the search space so analytic systems can focus on what matters. 

So where do models shine—and how do they strengthen, rather than replace, that human process? AI is built for high throughput analysis: it can sweep through years of longitudinal data, deduplicate and reconcile conflicting entries, and surface a ranked, concise view of what is new, what is unstable, and what truly requires a decision. The human brain, running on only about 20 watts, can efficiently apply empathy, ethics, and context to distilled analyses. The most powerful configuration is not AI or people alone, but the combination: models doing the exhaustive pattern finding and clinicians using that work as input to richer judgment about what should be done. 

The most powerful configuration is not AI or people alone, but the combination: models doing the exhaustive pattern finding and clinicians using that work as input to richer judgment about what should be done. 

In terms of application for best impact, acute care will always matter, but the long-term trajectory of health—and of health spending—will be shaped by post-acute and nonacute care, including home health, skilled nursing, assisted living, and hospice. Evidence from  hospital at home programs and home-based post-acute models shows that when clinically appropriate patients are treated outside traditional inpatient units, total episode costs can be thousands of dollars lower per episode, with a nearly 20% reduction in readmissions, lower complications, and higher patient and family satisfaction. Patients also spend more days alive and at home instead of in institutions——a metric that captures both better outcomes and what most people actually want from care. 

AI native infrastructure is essential: enterprise-wide platforms that plug directly into, or replace, legacy EMRs, automate interdisciplinary team reporting, and generate customer specific documentation in the background. When capabilities are delivered as core systems rather than isolated tools, organizations deploy across their full patient census, coexist with legacy records or replace them where appropriate, and scale without retraining the entire workforce. For a sector managing millions of longitudinal episodes under tight regulatory and financial constraints, the key need is precisely this: durable, AI enabled infrastructure that matches intensity of technology to intensity of care. 

In a well-designed post-acute system, AI sits in the background: orchestrating communication between teams, updating care plans, watching for early signs of deterioration, and ensuring that nothing important falls through the cracks. Humans remain at the center: assessing tradeoffs, discussing goals, building trust, and supporting patients and families through long arcs of recovery or decline. The technology increases the reach and reliability of care; it does not define its purpose. 

In a well-designed post-acute system...[AI] technology increases the reach and reliability of care; it does not define its purpose. 

When we commit computational and energy resources toward concrete, near-term objectives—fewer preventable hospitalizations, lower per episode cost, more time at home—we gain orders of magnitude more immediate value than chasing speculative “superintelligence,” in ways that are directly measurable in-patient outcomes and system efficiency. 

I write this as someone who has already had opportunities to build, lead, and learn across multiple chapters of my career—my team chose this work because it can support and protect what is irreducibly human in care. When we get the systems right—when we use AI to carry more of the work—clinicians, patients, and families have a better chance to bring what only they can bring: empathy in the room, ethics in the decision, and love in the long, hard parts of recovery and decline. 

The real promise of AI in healthcare is not that it will become human. It will deliver care that follows people over time instead of only meeting them in crises, give clinicians the focus to see and hear their patients clearly, and to support them in the places they spend almost all of their lives.