Connect for Patients (C4P)
A B2B healthcare product that helps clinicians find and verify relevant patient information more efficiently. AI surfaces useful context and supporting evidence, while clinicians retain control over interpretation and action.
ROLE
Solo Product Designer: product framing, interaction design, rapid prototyping, validation
Cross-functional product and engineering team
~ 2 months
Desktop and Web
Clinical information can be scattered across multiple sources, making it difficult for clinicians to quickly understand what matters for a patient.
The design opportunity was not simply to put more information on one screen. It was to reduce the work required to find, understand, and verify relevant context without introducing another system clinicians had to manage.
The challenge was to make AI genuinely helpful without making it feel opaque or overly autonomous. The experience needed to reduce search effort while preserving evidence, context, and clinician authority.
The goal was not to ask clinicians to trust AI more. It was to design an experience that gave them better reasons to trust, or question, what they were seeing.
I built a working high-fidelity prototype using AI-assisted development to explore the product direction beyond static screens. The prototype allowed me to examine information hierarchy, interaction states, progressive disclosure, and the handoff between AI assistance and clinician decision-making. It also created a shared artifact that could be discussed and evaluated across design, product, engineering, and clinician feedback sessions.
The first direction stacked EMR sections with Orion as a side chat panel: a sidecar, waiting to be asked. A nicer place to look at data, not a tool that did the looking. If Orion brings content forward, the organizing principle shifts from where things live to when things appear. So, Orion shows up three ways:
A system command
A shortcut
An always-there dock.
A 3-column stepped masonry with fixed widget sizes (XS to XL). The system keeps it full, so the provider curates by pinning, never builds from blank. Every widget shares one template: header, value, trend, provenance chips, "Ask Orion." That's what lets Orion compose the desk, any answer becomes a widget with "Add to desk," consistent with the rest.
Providers signed on as partners
Clinician research demos
Concept to fully functional prototype
This work reinforced that designing an AI-native product is not primarily about making AI more visible. It is about defining the right relationship between the user and the system.
AI should handle complexity where it can, but its behavior should remain clear, expected, and correctable. The user may not control every step, but they should retain meaningful control and final authority.
The strongest AI experiences give attention back to the work that matters.
