03 / Diagnoss
Diagnoss was designed as an EHR-embedded assistant that helps healthcare providers analyze clinical documentation, identify coding opportunities, and improve documentation without forcing them to leave the system they already use.
Role
Timeline
Platforms
Focus Areas
01
Overview
Diagnoss is an AI-assisted healthcare workflow designed to help providers document diagnoses, identify appropriate billing codes, and improve clinical documentation directly within their existing EHR environment.
The product evolved from a standalone healthcare application concept into a lightweight EHR-embedded assistant. Rather than asking providers to open another application, navigate between systems, or manually transfer information, Diagnoss brings intelligence directly into the clinical workflow.
The experience analyzes the patient's clinical documentation and provides contextual assistance while the provider continues working inside the EHR.
This evolution shifted the UX challenge from designing an application to designing an intelligent layer within an existing application.
Three primary capabilities
02
The Challenge
Medical documentation and coding are two of the most time-consuming parts of a provider's day. Notes must be complete, codes must be accurate, and both must align with the clinical record. Yet most providers still move between the EHR, coding references, and external tools to close the gap, slowing the encounter and increasing the risk of missing revenue or documentation.
Not this question
"How do we build a faster documentation tool?"
The actual question
"How do we streamline and optimize coding and note-taking inside the provider's existing workflow?"
The design needed to
03
Research & Discovery
Discovery combined conversations with clinical and business stakeholders, close reading of real documentation workflows, and analysis of where existing tools created friction. The central research question shaped every decision that followed:
“Where can intelligent assistance be introduced without interrupting the provider's existing workflow?”
| # | Activity | What it informed |
|---|---|---|
| 01 | Stakeholder interviews | Aligning clinical, product, and engineering perspectives on where assistance would be welcome and where it would intrude. |
| 02 | Workflow analysis | Mapping how providers move through chart review, documentation, coding, and sign-off inside the EHR. |
| 03 | Clinical documentation analysis | Reading real note structures to understand what information is present, implied, or missing. |
| 04 | User-centered design | Framing every feature around the provider's task rather than the model's capability. |
| 05 | Usability research | Iterative evaluations of the sidebar states, recommendation cards, and review actions. |
| 06 | Requirements gathering | Defining scope, compliance considerations, and EHR integration constraints. |
| 07 | Competitive / product analysis | Reviewing adjacent documentation and coding tools to identify patterns worth adopting or avoiding. |
| 08 | Cross-functional collaboration | Working closely with engineering and clinical stakeholders through design and implementation. |
Primary user
Needs
Pain points
Secondary stakeholders
Needs
Pain points
04
Key Findings
Providers already worked across complex systems and workflows.
Opening another application required providers to leave their current clinical workflow.
Recommendations become more useful when they understand the patient, encounter, and clinical note being reviewed.
AI recommendations needed to be understandable, explainable, and reviewable rather than presented as unexplained decisions.
“Don't make providers go to the AI.”
“Bring the AI into the workflow.”
The core product decision was to transform Diagnoss from a standalone destination into an embedded assistant. The EHR remains the primary workspace.
Diagnoss becomes a contextual intelligence layer that can be opened when needed, analyze information already present, and provide assistance without forcing the provider to leave the current workflow.
Old model — standalone application
New model — embedded sidebar
Artifact
10Before / after context-switching workflow
Seven steps across two systems collapse into a single continuous workflow inside the EHR.
05
Product Architecture
Existing EHR
The system of record
Diagnoss embedded sidebar
Diagnoss enhances the existing EHR rather than replacing it. Every input it uses already lives in the record; every output returns to the provider as a reviewable recommendation.
Artifact
09EHR + Diagnoss architecture diagram
How the sidebar reads from the active EHR context and returns recommendations into the provider's review.
06
The Product
The sidebar is organized around three primary stages. Together they create a simple mental model for how the system works — establish context, understand the note, then assist.
07
State contents
UX rationale
Before providing recommendations, the provider needs confidence that Diagnoss is connected to the correct EHR and patient context. This creates trust before AI assistance begins. Removing manual patient lookup reduces unnecessary steps and helps prevent workflow disruption.
08
State contents
UX rationale
Diagnoss analyzes the clinical note in context rather than treating the provider's input as an isolated text prompt. The system should make its analysis visible rather than presenting recommendations as unexplained AI output.
Detected entities
09
10
Marketing Site
Alongside the EHR assistant, Diagnoss needed a marketing site that could translate complex clinical AI capabilities into a clear, credible value proposition for providers, administrators, and health systems.
Outcome-first messaging
Revenue, accuracy, and provider happiness lead the story rather than technology claims.
Single primary CTA
A free coding assessment is repeated across the page so the next step is always obvious.
Healthcare credibility
Clinical credibility is established through team credentials, investors, and a calm, professional visual tone.
11
Outcomes
Diagnoss brings assistance directly into the provider's existing environment.
The assistant uses the active patient and clinical documentation as context.
Recommendations include supporting information and rationale rather than unexplained outputs.
Providers remain responsible for reviewing and accepting recommendations.
The experience reduces unnecessary navigation and manual information transfer.
The embedded-sidebar approach creates a foundation for expanding AI assistance across additional EHR workflows.
12
Reflection
This project changed how I think about AI product design.
“The best AI experiences don't necessarily ask users to interact with AI. They make the user's existing workflow more intelligent.”