Stella
My Role
UX Designer & Researcher
Worked with: PMs, engineers, designers, directors, leadership
Timeline
June 2025 - August 2025
My Tasks
UX Storytelling
Envisioned future-state
XFN collaboration
Prototyping & development
Tools
Figma
Adobe Photoshop
Roo Code
OpenRoute
The Main Problem
Care managers don't receive necessary data on time, often waiting weeks or even months for key patient health data reports.
My Deliverables
How Did I Help?
1
Design
Owned the design for all 3 chatbot screens, focused on usability and integration with current design systems.
2
Leadership
Proactively generated ideas and led mini sprints with my PM and engineer to fast-track development and iteration.
3
Development
The tight timeline meant in order to ship on time, I aided my engineer with development through Roo Code and OpenRouter.
Research, Research, Research
Assessing the Root Causes of Delayed Care
Before beginning work on any project, I always start with questions:
How do care managers currently request, receive, and use patient data in their daily workflow?
What barriers currently exist that prevent care managers from accessing data? Who do they rely on for data?
How does delayed or inaccurate data affect care managers’ ability to support patients
Research Methods
Analyzing previous care manager research done for Care Connect, included analyzing historical interview data, workflow studies, and pre- and post-launch personas. I focused on understanding internal workflows, data access barriers, and system usability challenges unique to Blue Shield’s care managers.
Partnering with engineering to understand how data is structured, accessed and managed at BSC.
Key Findings
Delayed Decision Making
Care managers can't act quickly with the current system. By the time patient data arrives, patient needs may have already changed.
Middleman Dependency
Receiving data requires contacting someone with patient data access and waiting for a report, which causes numerous delays.
Inability to Drill Into Data
Data often arrives in unusable formats for deeper analysis, requiring additional data requests and continuing the cycle.
Delayed Decision Making
Patient health data is pulled from 13 different sources, making a long, complicated data access process even more problematic.
Primary Goal
Cutting Out the Middleman
Give care managers a way to access reliable patient data independently and in real time — reducing reliance on manual requests and shortening the gap that exists between needing information and being able to receive then subsequently act on it.
What this means: Create a system that allows care managers to independently query and retrieve patient data in real time.
My Hypothesis
Create a Single Source of Truth
Leadership had aligned on the problem, but not the solution. How we got there was left entirely up to our intern team. Once I had clarity on the goal, I hypothesized that a generative AI chatbot, pulling from a centralized database through natural-language questions, would allow care managers to get accurate PHI answers back immediately, without waiting on a manual request. This idea became the concept we built and tested with Stella.
Meet Stella
*Screens partially blurred to protect NDA, No real patient health information is shown here. Data was drawn from fabricated PHI in a JSON document.
Stella in Action
To prove Stella's value beyond the interface design, I worked with my UX team and a care manager to draft real scenarios a care manager might encounter in their day-to-day work that would have real impact on BSC's patient outcomes. Below is an example of Stella responding to one of these scenarios:
87%
Success Rate. Stella successfully handed 13/15 use cases.
Stella's Impact
Informing the Future of Data at Blue Shield
Stella was developed as a proof of concept project, meant to validate my hypothesis that use of a GenAI chatbot to query a centralized data base would allow care managers to work more effectively and efficiently. The prototype demonstrated clear business value by showing care managers could retrieve accurate data both more quickly and independently.
Validating Business Viability
Stella was technically feasible and proved to have both user and business value by showing that it could create meaningful reductions in the time it takes to get data. This confirmation gave Blue Shields’s leadership confidence to commit to larger-scale investments in similar tools to solve the data problem, one of which is currently being developed.
Guiding Future Decision Making
As the first tool of its kind at BSC, Stella became a reference point for shaping future internal data products. Its success helped answer key strategic questions, confirmed that care managers would adopt this approach, and now Stella informs the development of a new enterprise data tool built on the same foundational concept.
*Mockup of Stella integrated into a current Blue Shield of California product used to prove integration feasibility with existing systems.
Navigating Challenges
A Not So Smooth Sea
The path to creating Stella was NOT straightforward. Early in the project lifecycle, our team faced significant ambiguity around what the project goals were. Requirements shifted weekly, sometimes even multiple times within the same week, as leadership aligned on priorities for our project. Our first concept was not remotely similar to what Stella ended up becoming, it wasn’t even in the field of care management. This created numerous challenges, namely:
Unclear Scope and Direction
To manage unclear scope, I focused the team on features that supported the core concept and held frequent alignment sessions across our UX, engineering, and PM teams to quickly assess feasibility and pivot as priorities shifted, which prevented scope creep and saved significant time.
Compressed Timeline
With only 3 weeks from conceptualization to presentation to the C suite, I drove a fast, focused design process. Prioritizing essential functionality and rapid iteration kept the team moving. I also learned OpenRouter and Roo Code to directly support development alongside our engineer rather than waiting on handoffs.
Stakeholder Communication
To navigate the lack of clarity, I led close, ongoing communication across my UX team to eliminate redundant work, and proactively escalated ambiguity to my manager when it blocked progress, driving the alignment leadership needed to clarify our direction.
Reflection and Learnings
Adapting to Change
My internship at Blue Shield of California can be summed up by the word adaptability. I had to stay agile as project scope shifted multiple times, rethink work that had already been started, pivot frequently and apply tips and tricks from my incredible team taught me on the fly. This pace pushed me to learn new tools and adjust my usual workflow to keep the project moving forwards.
1
Advocating for Design
Working through ongoing ambiguity taught me how to defend design decisions with research and user needs, especially when leadership priorities shifted. I learned to communicate the “why” behind design choices and that helped better align cross-functional partners.
2
The Power of Communication
Confusion in the project made it clear how essential proactive communication is. Once I began sharing frequent updates with all project stakeholders, clarifying what needed to be designed and looping in my manager to help translate leadership expectations, the project moved forwards at a steady face even in the face of shifting requirements.
3
Empathy and Shared Learning
Collaborating with engineering and care management showed me the power of empathy and knowledge-sharing. By exchanging skills through design workshop sessions and technical walkthroughs, we strengthened our partnership as well as improved the clarity, quality and development of the product.
Manager Feedback
POV: Working with Danial
