AI COST OPTIMIZATION
Designed a predictive insights dashboard that helps users explore large datasets with AI-assisted summaries and recommendations.
THE CHALLENGE
Users needed a way to understand trends and forecast cost outcomes associated with shift pricing at their medical facilities without relying on manual analysis or exporting to external tools. The goal was to build an interface that complemented user expertise with machine-generated insights, enabling faster, more confident decisions.
ROLE & SCOPE
I served as the lead designer, owning research planning, concept development, UX design, interaction patterns, and collaborative reviews with engineering and product leadership. I worked closely with data science teams to understand model outputs and translate them into human-centered experiences. My role included shaping how predictive insights are surfaced, structured, and acted on.
RESEARCH & INSIGHTS
Research through user flows and stakeholder interviews highlighted two major user needs:
​
• Contextualization of AI predictions alongside historical data
• Control over how and when AI summaries are presented
​
Based on this, I defined design requirements that ensured AI insights felt trustworthy, transparent, and actionable.
USABILITY TESTING & DECISIONS
After building a prototype and conducting usability testing, key decisions included:
​
• Crafting visual patterns that place predictive insights near relevant performance metrics
• Designing contextual help and explanation layers for AI outputs to build understanding and trust
• Creating consistent states for forecast scenarios so users could easily compare outcomes
• Structuring the dashboard to support iterative exploration of recommended actions
​
These strategies balanced automation with user agency, a critical consideration for enterprise workflows.
OUTCOME & IMPACTS
After synthesizing usability findings, I translated the improved flows into high-fidelity designs supported by thorough documentation for developer handoff. I partnered closely with engineering throughout implementation, ensuring design fidelity and accessibility standards were maintained. This continuous collaboration allowed the product to move efficiently from prototype to production.
​
This dashboard allows users to:
​
• Quickly interpret forecast scenarios without manual modeling
• Explore recommendation pathways tied directly to underlying data
• Trust predictions through contextual transparency and explanation layers
​
This design contributed to greater user confidence and reduced time spent on data interpretation tasks.
