CANCER COMPASS
Empowering people with cancer to learn with confidence
A calmer way to learn after a cancer diagnosis
MY ROLE
Design Strategist
Product Designer
TIMELINE & STATUS
6 Months
Presented to leadership
After diagnosis, many patients free-fall into solo searching and medical jargon while clinicians backfill with corrections during valuable consultation time. With 2M+ new U.S. cases annually, health systems can’t guide everyone.
I led strategy, co-design, and prototyping to identify a patient-safe, scalable AI use case for Microsoft’s Customer Experience team and their clients.
Early signals are strong: patients described a calmer alternative with 100% reported trust in AI generated summaries, clinicians confirmed the gap it fills in the early stages, and Microsoft CX endorsed the concept for impactful use of AI for healthcare.
INTRODUCING
Cancer Compass: a personalized cancer education platform
to guide patients and their care team throughout the journey.
“This platform would’ve saved me hours of anxious googling and shown me information that is actually helpful.”
- Patient we interviewed
PROCCESS
Research and co-design rooted in a holistic understanding of the cancer journey
Multi-stakeholder mapping
We interviewed patients, caregivers, clinicians (doctors, nurses) and AI/accessibility experts to understand the whole picture. We also analyzed hundreds of real posts in cancer forums to triangulate patterns beyond interviews.
Trauma-informed, participatory practice
Sessions used progressive disclosure, neutral moderation, breaks, and flexible modes to reduce cognitive/emotional load
Experience-based co-design
We treated stakeholders as co-designers, mapping touch points from the post-diagnosis “limbo period,” then co-creating UI patterns that respond to lived experience
INFORMATION OVERLOAD
The post-diagnosis “limbo period” is high-need and high-risk
In interviews with 8 cancer patients, everyone named the post-diagnosis “limbo period” as the most overwhelming in the whole cancer journey, made worse today by unlimited, unfiltered access to articles, blogs, and forums. People are alone with the internet, spiraling down rabbit holes, unsure what applies to their case. Half spent hours manually filtering articles for trustworthiness and relevance, and all eight said the medical jargon felt like learning a new language.
Clinicians echoed the cost on their side: oncologist visits are spent debunking search-driven panic instead of moving care forward.
How might we make relevancy obvious and understanding simple?
KEY DECISIONS & FEATURES
Personalized guidance through trusted sources
1. EHR-aware filtering (opt-in)
Use key fields (diagnosis, stage, treatment) to rank pre-approved articles for this patient.
2. Guided research questions
Generate tailored questions that guide users to ease into learning about the nuances of their demographic or case
3. Why this is relevant” summaries
Summary of each peer reviewed source outlining the relevance to specific patient’s case.
Lower cognitive load to improve and build understanding
1. Adjustable reading level
Simplify summaries without losing fidelity to sources.
2. Tap-to-define terms inline
Stay in flow; no tab-hopping
3. Highlight to learn more
Expand sections for quick context and key takeaways.
CHALLENGE
AI can be a powerful tool for personalized learning, but there is low trust, especially for health
In formative sessions, participants recognized their own stories in the flow: more linear learning flow which is less overwhelming than have 100 tabs open, clearer “this is for me” signals, and language that met them where they were.
From a concept standpoint, all participants reported that this would be very useful if they had known about it when they were researching at the start of their journey. However, there were concerns around trusting AI and the content.
How might we make relevancy obvious and understanding simple?
BUILDING TRUST AROUND AI
Collaborative, transparent, and neutral - mitigating the potential risks and harm of AI
We worked with AI experts and healthcare professionals to shape design requirements for AI in healthcare and co designed the language with patients focusing on:
Augment, don’t replace
No medical authority or directives; suggestions only, with escalation to providers.
Radical transparency
Every AI-generated summary is labeled and tied to sources and dates. Doctor approved sources where you can trace back to the citation and the original article very easily
Neutral-positive tone
Plain neutral language with a hint of warmth. Not overly positive which takes away credibility, but also not devoid of all emotion, which can be lonely
After implementing changes to language and flow, we tested with 5 patients and 100% said that they found the platform and use of AI trustworthy.
HEALTHCARE AS A SERVICE
Prioritizing collaboration to support the realities of cancer care
While the most overwhelming time is being alone with the internet, caregivers an doctors we talked to revealed the holistic nature of cancer care. Mapping out the caregiver and doctor experiences helped us identify points in the experience that would significantly enhance the patient experience.
Caregiver sharing
Send plain-language summaries so loved ones stay informed without burdening the patient to align with the reality of support systems.
Visit prep
Capture questions as you read; share with your care team ahead of appointments to improve appointment quality and preparedness.
Clear demand, better prep, higher trust—ready for client demos
Patients described the experience as calming and asked when they could use it. 100% of patients we tested with reported that they would trust the content being generated.
Clinicians said it fills a real capacity gap in the post-diagnosis window, possibly helping patients navigate the in between windows when they don’t yet have access to a medical professional are are in mental distress.
Microsoft CX team endorsed the direction and research for pitching AI use cases to their healthcare clients. Praised the storytelling, pain pont articulation, scale and gap, and case for leveraging the power of AI in this space.
“This work pinpoints a clear emotional use case for AI in healthcare and the team really worked to defined how AI can show up safely in this space.”
- Microsoft CX lead researcher
CONCLUSION & LEARNINGS
Co-creating safe and useful AI tools
Designing for AI
AI can be a very powerful tool, but it bust be designed with extra caution to ensure safety and it something that is truly useful rather than harmful.
Mixed methods research for sensitive topics
Using techniques like progressive disclosure, semistructured interviews and supplementing with netnography and survey to validate findings is useful for working with sensitive populations.
Service Design
Healthcare is a network—design for patients, caregivers, and clinicians together. If we had more time, I would do more research on the backstage actions and tools in hospitals
