case study, 2026
project Summary
To design an AI clinical Copilot, prioritizing patient-centered outcomes through the strategic application of AI. By moving away from retrospective manual audits, this intervention provides real-time feedback for physicians, ensuring that every heart failure patient receives the highest standard of care during their hospitalization.
Industry
location
Heathcare / Medtech
Toronto, Ontario, Canada
Client
Tools
Sinai Health Foundation
Figma (Advanced Prototyping & Variables), LLM Logic Mapping
Physicians-side
A high-density, dynamic dashboard provides a macro-view of the entire inpatient roster. It highlights care gaps in real-time, allowing doctors to optimize therapy immediately rather than waiting for post-discharge reviews.
The Goal
Physicians need to know which patients need their attention right now and what exactly they should do in under 10 seconds
Pain Points
Fragmented Clinical Information
Critical data is scattered across the EHR (notes, labs, orders), requiring constant context-switching within systems. This slows down decision-making and increases the risk of oversight.
Missed or Delayed Interventions
Eligible treatments (e.g., medications, diagnostics) are often delayed or missed due to time pressure and lack of real-time visibility.
Cognitive Overload During Rounds
Currently, QIs are checked manually via "chart-mining" - a fragmented process that takes 5–10 minutes per patient - leading to decision fatigue and inconsistent care.
Retrospective Feedback Loops
Current quality monitoring relies on audits after discharge, making it impossible to correct missed care in real time.
problem
Despite robust evidence-based guidelines for Heart Failure (HF) care, hospital physicians often struggle with “information overload.” Critical Quality Indicators (QIs)—such as the timely initiation of GDMT (Guideline-Directed Medical Therapy)—are frequently missed due to fragmented data across EHR systems. Traditional audit-and-feedback happens weeks after discharge, when it is too late to impact the patient’s immediate outcome.
solution
Utilize an LLM to scan clinical documentation, lab results, and medication orders continuously. It measures these against the EQUAL-HF quality indicators (e.g., ACE inhibitor prescription, echocardiogram completion, and discharge education).
wireframes
Physician User Experience is: Dense, Fast, Action-driven, Clinical language
Design System
Physician User Experience is: Dense, Fast, Action-driven, clear.
typography
Geometric but soft font that feels “clinical-tech”, but not cold, and has excellent readability and works well in dense data environments (EHR-style UI)
color
I developed a clean, healthcare-appropriate palette that uses blue + accent blue to feel like a modern healthtech AI product and make the alert color system stand out
Primary Brand
HEX: 5D5EDA
Secondary Brand
HEX: 4FD1C5
Neutral White
HEX: E8E8E8
Neutral Black
HEX: 1B1E27
final UI
conclution
The success of this feature highlighted the importance of early stakeholder alignment, iterative design, and designing for real-time decision-making in high-stakes environments.
next steps
By focusing on real-time visibility and efficient response actions, we significantly enhanced the effectiveness of the system, making medical teams more proactive and responsive. Lessons learned here will inform future improvements in live monitoring and automation.
Ready to connect?