CASE STUDY
MEDSCRIBE AI
Healthcare providers spend a significant portion of their time documenting patient visits instead of treating patients. AI has the potential to assist with clinical documentation, but must be implemented carefully to preserve physician control, patient consent, and data security.
This project explores how AI can augment workflows for medical professionals by assisting with clinical charting.
Using modern AI tools, I built a functional prototype of an AI-powered medical scribe and tested whether AI could meaningfully accelerate the product design workflow itself. Intentionally design for desktop only so that data doesn't travel outside of a controlled network on a mobile device.
The Problem
Clinical charting requirements create several issues:
• reduced patient interaction
• physician burnout
• delayed documentation
• administrative overhead
• reduced patient interaction
• physician burnout
• delayed documentation
• administrative overhead
Medical scribes can help, but they are expensive and difficult to scale.
AI offers a potential solution, but healthcare introduces strict ethical and privacy requirements.
Results Achieved
4 hour prototype production
Reduced typical prototype development time from ~2 weeks to 4 hours.
Reduced typical prototype development time from ~2 weeks to 4 hours.
AI-assisted design pipeline
Used Claude, Claude Code, and Sigma to generate and convert designs into editable UI prototypes.
Used Claude, Claude Code, and Sigma to generate and convert designs into editable UI prototypes.
Ethics-first interaction model
Designed explicit patient consent flows and physician sign-off safeguards.
Designed explicit patient consent flows and physician sign-off safeguards.
Production-ready structure
Figma components (including Auto-layout) created using Figma Make. Streamlined engineer handoff with potential to improve cross functional collaboration.
Figma components (including Auto-layout) created using Figma Make. Streamlined engineer handoff with potential to improve cross functional collaboration.
My Role: Product Designer, UX Researcher, AI Workflow Architect
Objectives
• Product concept development
• User research interviews
• Ethical design framework
• UX architecture
• AI-assisted prototyping workflow
• Interaction design
• Product concept development
• User research interviews
• Ethical design framework
• UX architecture
• AI-assisted prototyping workflow
• Interaction design
The AI WORKFLOW EXPERIMENT
The main goal of the project was to test how AI could accelerate the design process itself.
I experimented with a workflow using:
• Claude for concept ideation and UX logic
• Claude Code for generating front-end code
• Figma Make to convert the code into editable design components in Figma
• Claude Code for generating front-end code
• Figma Make to convert the code into editable design components in Figma
Design Princples
Through conversations with physicians and healthcare professionals, three priorities emerged.
Physician Control
Doctors must remain responsible for the final chart. AI can assist drafting notes, but physicians must review and approve documentation before it becomes part of the medical record.
Doctors must remain responsible for the final chart. AI can assist drafting notes, but physicians must review and approve documentation before it becomes part of the medical record.
Patient Transparency
Patients must know when AI is involved in documentation.The system requires clear consent before recording or generating notes from a consultation.
Patients must know when AI is involved in documentation.The system requires clear consent before recording or generating notes from a consultation.
Data Security
Healthcare documentation requires strict compliance with privacy regulations.The design prioritizes minimal data exposure, clear storage boundaries, and physician oversight before record submission
Healthcare documentation requires strict compliance with privacy regulations.The design prioritizes minimal data exposure, clear storage boundaries, and physician oversight before record submission
The interface was designed with accessibility in mind.
Key design choices included:
• dark mode by default to reduce eye strain
• breadcrumb navigation to support cognitive clarity and focus.
• minimal UI noise to support focused documentation
• 8-point grid system for consistent layout
• These decisions help maintain low cognitive load during clinical workflows.
• breadcrumb navigation to support cognitive clarity and focus.
• minimal UI noise to support focused documentation
• 8-point grid system for consistent layout
• These decisions help maintain low cognitive load during clinical workflows.
RESULTS
The prototype successfully demonstrated that AI can significantly accelerate early-stage product design.
Key takeaways:
• AI dramatically speeds prototype production
• Human designers remain essential for strategy and ethics as well as arbiters of what solutions best fits the problem–this has always been the fundamental purview of the designer and remains the case in the AI era.
• AI works best as a design accelerator, not a replacement
• This approach also improves the prototype and the developer handoff process, since the prototype already includes structured code and component layouts. However the production in the middle and creative decision is best done by human designers.
The prototype successfully demonstrated that AI can significantly accelerate early-stage product design.
Key takeaways:
• AI dramatically speeds prototype production
• Human designers remain essential for strategy and ethics as well as arbiters of what solutions best fits the problem–this has always been the fundamental purview of the designer and remains the case in the AI era.
• AI works best as a design accelerator, not a replacement
• This approach also improves the prototype and the developer handoff process, since the prototype already includes structured code and component layouts. However the production in the middle and creative decision is best done by human designers.
