"Pretty Prose" Alone Won't Fix Healthcare

Here’s How We Are Structuring Healthcare Data for Non-Acute Care

The first thing most people notice about our AI documentation is that the notes sound better. A LOT better. But our goal is not just better sounding notes. Our real job is building non-acute care data systems. These systems’ surfaces are beautiful, but their architectures—structured, trustworthy, and primed for action—can shoulder the weight of an enormous social need.

Chronic illnesses affect around 60% of Americans, who often have multiple conditions. Effective, timely outpatient care can reduce the need for hospital treatment. But the data and systems in non-acute care have not kept pace with the need.

My teams and I have always purposefully chosen problems like this to solve, and it has been a great privilege to do so, in many kinds of complex systems. A pathway conducts, or it does not; a cluster either carries signal, or it quietly blocks it. In batteries, we balanced conduction, diffusion, and mechanical stress to enable storage that could reliably support more efficient transportation. In biological materials—actin filaments carrying signaling ions along their length, mitochondrial cristae whose shapes tune electrochemical potential, collagen and other scaffolds that hold tissues together—the work showed, again and again, how precise, deliberate changes in structure can produce large, observable changes in function. The common method in all of these challenges? Start with a human problem and use mathematics to locate the point where strategic, structural improvements enable enormous gains in performance.

...the work showed, again and again, how precise, deliberate changes in structure can produce large, observable changes in function.

Non-acute care has presented the same kind of problem we have met in other complex systems: critical pathways are blocked. Consider the everyday reality for nurses, who spend, on average, around 40% of their working time on documentation and related electronic record tasks. This time directly penalizes the time that nurses have with patients. Time and motion studies show similar patterns; one analysis found nurses spend more time on electronic health record tasks than on direct patient care.

Meanwhile, U.S. health spending reached about 5.3 trillion dollars in 2024—18% of the entire economy—with spending growth outpacing economic growth. Much of this burden is driven by chronic disease and by the downstream costs of inadequate prevention and management. The costs show up on national balance sheets, and they also show up in family bank accounts, lost work, and exhausted caregivers.

In this context, AI that only produces nicer paragraphs or “pretty prose” is just cosmetic smoothing on a high noise, poorly controlled system. Healthcare needs structured data—captured once and reused many times—flowing through secure, AI-first architectures. It needs systems that represent vitals, medications, diagnoses, interventions, outcomes, time, and place as well-defined elements, not just as sentences.

...AI that only produces nicer paragraphs or “pretty prose” is just cosmetic smoothing on a high noise, poorly controlled system.

Nurses and caregivers in non-acute settings are central to this effort. They know whether a new medication is actually taken, whether a wound looks better, or whether a family can manage another week at home. They are the high degree nodes in the graph—the points where flows converge. We must make it easy and natural for healthcare professionals' everyday work to produce structured information as a graceful, intentional side effect—not as an extra job.

A focused technology company can play a constructive role here. Most care organizations cannot, on their own, design and operate modern AI infrastructure at the level of security, reliability, and performance that this problem demands. They are busy taking care of people. A dedicated technology team, working in partnership with clinicians, can build the underlying architectures: cloud based, secure, designed from the start for both machine reasoning and human use.

At Amesite, we built NurseMagic™ with that ambition. Our systems are designed so that normal use—documenting a visit, updating a plan, communicating with a team—create structured data: problems, medications, observations, interventions, outcomes, and context tied to time and place. They are designed for mobile, team-based, non-acute workflows, not just for office visits. And they are built for strong security and processing that emphasizes relevant information, rather than adding yet another layer of alerts.

We are turning notes from isolated clusters of text into connected pathways of data—streaming safely from the nurse at the bedside to the people and systems that need to act. When those pathways are in place, the same documentation that used to drain time can instead support earlier detection of risk, better staffing decisions, and clearer insights into care delivery.

Each constraint we remove is opening new channels in a crowded network: signals are traveling farther; patterns are coming into view, and the quiet, careful work of clinicians is gaining the support it has always deserved. It is the kind of problem that attracts remarkable people, and I am grateful every day for my colleagues at Amesite, who bring their intelligence, skill, curiosity, and goodwill to it. We build for providers, so that the people doing the hardest work have tools that finally match the importance of what they do.

We build for providers, so that the people doing the hardest work have tools that finally match the importance of what they do.

Our team feels a strong responsibility to deliver these systems, and we chose this work deliberately. The need is urgent, the impact is broad, and the skills we have built in other complex systems have enabled us to make fast progress with live systems and deliver measurable outcomes. In the coming months, I plan to write here regularly about what we see in real deployments for care professionals and organizations. We invite people working in non-acute care—nurses, aides, therapists, managers, physicians, and policymakers—to read, respond, and help us shape the next steps of this technology.