Transforming raw data into actionable insights for healthcare CFOs
Repeated data Dashboard Use
Insight cards Interaction
Nov. 2024 - May. 2025
1 Product Designer
1 Developer Lead
2 Developers
2 Quality Managers
Lead product designer from concept research to product final launch
Customer Interviews, Prototyping, Usability Testing, Iterative Design, Product Thinking
OVERVIEW
Epic Systems is a leading healthcare software provider, specializing in electronic health records (EHR) and integrated health information systems. Its solutions help hospitals and clinics manage patient care, clinical operations, and administrative tasks seamlessly.
Instead of relying solely on open-ended AI chats, my approach leveraged clear, actionable insight cards paired with essential manual controls tailored for experienced users.
INITIAL PROBLEMS
Financial Pulse is a comprehensive benchmarking dashboard for healthcare CFOs. Despite its rich data, busy executives found the experience overwhelming, leading to low adoption after initial use.
As the lead product designer, I found that the issue wasn’t the lack of data—it was the lack of clarity, direction, and confidence. Despite the dashboard’s depth, users felt overwhelmed rather than empowered.
EARLY RESEARCH
We started conducing user interviews and talk with end users during immersion, and make journey map to understand their workflows.
KEY INSIGHTS & GOALS
Identify key metrics quicker without scanning metrics.
Increase return usage and more engage using frontier AI technology.
PROBLEM STATEMENTS
DESIGN CONSTRAINTS
Denial Pulse is just one of six Pulse modules in the Financial Pulse platform. To maintain consistency and avoid feature disparity across modules, we could not redesign the main dashboard layout — including the central 20+ key metric cards.
Our solution had to layer clarity and guidance on top of the existing structure, without disrupting shared components or creating feature inequality across Pulse modules.
ITERATION #1
We found out the current pain points are users always scanned data, but no idea was important, so we introduced a conversational AI chatbot to help users explore metrics by asking questions. The idea was to reduce cognitive load and offer clarity through an intelligent assistant.
However, it quickly failed. For busy CFOs, this added complexity rather than simplifying their workflow. The approach mismatched their needs, leading us to rethink our AI strategy entirely.
Our research led us to ask: Under the age of AI, is AI chatbot always the right answer of everything?
We reframed our goal from designing "AI" to designing clarity. We created structured, rule-based insight cards that surfaced top issues, suggested next steps, and explained potential impact.
ITERATION #2
We initially translated all necessary information into numbers, but user feedback revealed frequent misinterpretations of the data’s meaning. In response, we refined the card-level design to strike a better balance—surfacing key metrics up front while allowing users to expand collapsed sections for more detailed information.
We keep the data table because of their muscle memory and workflows. AI existed as guidance, not gatekeeping.
FINAL DESIGNS
We refined the card-level design to strike a better balance—surfacing key metrics up front while allowing users to expand collapsed sections for more detailed information.
Actionable Insights Cards
Deliver quick, clear, and practical recommendations at the top of the dashboard.
Explainable AI Chatbot
Provide an interactive chatbot that lets users explore insights in depth
REAL IMPACT & RECOGNITION
The redesign received strong positive feedback from users. CFOs found the new insight cards and trend overlays surfaced what mattered most—without extra effort or training. They felt more confident in their decisions and saved significant time during weekly reviews.
Repeated data Dashboard Use
Insight cards Interactio
We shared our future directions at Epic’s major event, XGM, where the project was showcased as a leading example of how AI can drive clarity and usability without overwhelming users.
REFLECTION
Open-ended prompts created friction for users under pressure. They didn’t need a chatbot—they needed clear, prioritized guidance. Also, experienced users value control. Designing for augmented intelligence, not automated intelligence, led to higher adoption and satisfaction.
