There's a particular failure mode I've watched play out with nearly every attempt at pre-visit chart review, including some very expensive ones. The tool dutifully pulls everything it can find about the patient — every lab, every note, every prior encounter — and presents it. Comprehensively. Exhaustively. Uselessly.
The clinician opens it, sees a wall of information, and closes it. Then they do what they always did: skim the outside records by hand during the visit. The expensive tool changed nothing, because it solved the wrong problem. It optimized for completeness when the constraint was attention.
The constraint is attention, not data
A pre-visit summary lives or dies on a brutal economic fact: the clinician has somewhere between thirty seconds and two minutes to read it before the patient is in the room. Any design that requires more time than that will not be used, no matter how complete it is.
So the design problem isn't "how do we show everything?" It's "what are the three to five things this clinician needs to know before they walk in, for this patient, at this visit?"
A good pre-visit summary is defined by what it leaves out. Completeness is the enemy of usefulness when the reader has ninety seconds.
Three things a useful summary does
1. It leads with what changed
The single most valuable thing a returning clinician needs is the delta. Not the patient's entire history — they have a sense of that — but what's new since they last looked. A new diagnosis from an outside specialist. A medication the cardiologist adjusted. A lab trend crossing a threshold. An ER visit nobody told them about. Lead with the delta and you've earned the clinician's attention for everything else.
2. It collapses, then expands on demand
The summary should fit on one screen in its collapsed form — the headlines a clinician can absorb in a glance. But every line should expand. When the clinician's eye snags on "medication adjusted," one click shows the detail, and one more shows the source document. Progressive disclosure respects the ninety-second budget while keeping the full record one tap away.
3. It stays traceable
This is non-negotiable, and it's where AI-generated summaries earn or lose a clinician's trust permanently. Every claim in the summary must point back to the source document it came from. A summary the clinician can't verify is a summary the clinician won't rely on — and rightly so. The value isn't the AI's confidence. It's the clinician's ability to check.
The best pre-visit summary isn't the most complete one. It's the one a busy clinician actually reads — and trusts enough to act on.
Why this is a design problem, not an AI problem
It's tempting to think the bottleneck is model quality — that a better model would produce a better summary. But the models are already good enough to extract and structure clinical information well. The thing that's usually broken is the product decision about what to surface, in what order, at what density.
That's a design discipline, and it's one most healthcare AI hasn't internalized yet. The instinct is to demonstrate capability by showing everything the model found. The discipline is to hide most of it, and surface only what serves the next decision.
How we think about it at MediClarity
We design the pre-visit summary backward from the visit. The collapsed view answers "what changed and what should I pay attention to?" in the time it takes to walk down a hallway. Every line expands to detail, then to source. Connected wearable data and outside records fold into the same view, so there's one place to look, not four.
It's deliberately less than what we could show. That restraint is the feature. If you want to see what that looks like on a real chart, the demo walks through a sample patient and one of yours.
Dr. Daniel Korya is a board-certified vascular neurologist and the founder of MediClarity. This piece reflects product-design opinion informed by clinical practice, not a claim about specific time-saved outcomes.