Embrace Health AI

Duration

04/2023 to Present

Deliverable(s)

Hi-Fi Prototypes

My Role

UX Research
User Testing

Product Strategy
UX/UI Design

Collaboration

Care Providers
User Researcher
ML Engineer

Sharman
Behavioral Researcher

Challenges

"When your pain repeats, do you need therapy?"

"How could my procrastination be tied to…a lack of love?"

In 2022, I worked on a design project supporting individuals who had lost loved ones. When I felt myself lucky to have my fam healthy, a year later, after losing my grandfather, I experienced not only the sorrow but also the grueling wait times for mental health support firsthand.

In 2022, I worked on a design project supporting individuals who had lost loved ones. A year later, after losing my grandfather, I experienced not only the sorrow but also the grueling wait times for mental health support firsthand.

People often overlook lingering negative emotions—until a life event triggers a deeper awareness. It starts with noticing small, recurring moments of pain. Over time, these patterns become impossible to ignore, prompting the question: Why do these painful experiences keep repeating? This is where the need for mental healthcare begins.

Yet despite growing mental health awareness, cultural taboos and systemic barriers still prevent timely care. In the U.S., patients wait an average of 12 weeks for therapy and often drop out before their first session due to a lack of interim support.

Care providers also face challenges: helping patients open up without retraumatizing them, especially in diverse communities. For example, providers hesitate to offer self-help resources to blue-collar patients, who may have lower literacy or less trust in formalized care. Without tailored guidance, these patients are often left behind, trapped in cycles of unresolved suffering.

Roles & Goals

As the end-to-end product designer, I aimed to:

Identify the barriers patients and care providers face during initial therapy.

Research and design effective, user-centered mental health solutions in reshaping mental healthcare practices through behavioral studies.
Prototype emerging technologies like AI to explore their real-world application in mental healthcare.

Research & Design

Through behavioral studies back to 2022, I gained valuable insights from multiple rounds of research:

Journaling & Storytelling

Digital journaling tests showed users resonate more with dramatic stories that mirror their own struggles. While trauma survivors fear revisiting painful experiences, they can be empowered and inspired with the right prompts.

I also found that reframing goals in the past tense ( e.g., “I have finished...” vs. “I will...”) led to higher motivation, which is scientifically evidenced about sub-conscience.

Gamification for Engagement

Educational prototypes revealed that users seeking mental healthcare respond poorly to vague or general questions but engage more when tasks feel purpose-driven and explanations are transparent.

Customization for Accessibility

Care providers noted that pre-made therapy content often fails to meet the needs of blue-collar groups, highlighting the need for adaptive, customizable and self-paced approaches.

Balancing AI & Human-Centered Care

While care providers were cautious about AI’s role in therapy, they welcomed AI’s ability to automate labor-intensive tasks like documentation and smarter triage.

The How-Might-We Guide emerged with a central question:

Key findings from interviewing 70+ designers in the home decor industry:

How might we support mental healthcare patients in their daily routines so therapists can better understand them and speed up their therapy progress?

Figure: Conversational AI-powered patient management portal for care providers

Figure: A typical gamified flow of learning humans' train of thougts by AI parsing & NLP

Figure: A AI-powered customized triage process to better serve patients' needs

Figure: Converational AI helping patients digest protocol-rich documents

Figure: Content rewriting (e.g., showing expected durations of tasks, enabling interactions between care team & patients)

to enhance patient engagement

Strategic Collaboration

MVP Dev.: Given that this innovative project operates within a regulation-heavy industry and that current generative AI models remain unpredictable for real-world testing in such a sensitive domain, I broke down ambitious concepts into smaller, testable components to enable faster prototyping.
For example, when designing triage questions to understand people’s mental stableness, instead of building a large language model to ask random testing questions, I tested with engineers merely on emotion-related questions that categorized users into types.

MVP Dev.: Given that this innovative project operates within a regulation-heavy industry and that current generative AI models remain unpredictable for real-world testing in such a sensitive domain, I broke down ambitious concepts into smaller, testable components to enable faster prototyping.

For example, when designing triage questions to understand people’s mental stableness, instead of building a large language model to ask random testing questions, I tested with engineers merely on emotion-related questions that categorized users into types.

Prototype Roadmap: I led iterative design reviews with mid-fi and hi-fi prototypes to keep stakeholders aligned and manage expectations around early-stage concepts versus final outputs.

Prototype Roadmap: I led iterative design reviews with mid-fi and hi-fi prototypes to keep stakeholders aligned and manage expectations around early-stage concepts versus final outputs.

Coding an ML model for a more intuitive user flow of emotion recognition

Business Impact

$2.5K

Funding for research

My team secured through a top-three finish in an entrepreneurial competition, attracting VC interest for further research.

28–30%

Increase in Completion Rate

From the very first journaling prototype to the latest AI-powered solutions for both care providers and patients, a survey showed more completion of journaling gaming.

Learning

This is the project where I gained more learning than impact:

Preparedness for Sensitive Research: Mental health research requires backup plans for emotional breakdowns and disruptions. Standard workflows may need to be paused or adapted in real time.

Designing for Trust in AI: Designing AI features for high-stakes scenarios like mental healthcare requires meticulous consideration. It’s crucial to set clear expectations for users and offer a sense of control. Building trust in AI takes time and should be approached through digestible, bite-sized guidance—such as tooltips or an always-on AI assistant to navigate resources—rather than a rigid step-by-step onboarding process.

Iterative Mindset: Working with engineers on a complex, industry-wide challenge reinforced the importance of iterative collaboration to deliver an MVP with core features, while saving "nice-to-have" features for future iterations.

Designing for Trust in AI: Designing AI features for high-stakes scenarios like mental healthcare requires meticulous consideration. It’s crucial to set clear expectations for users and offer a sense of control. Building trust in AI takes time and should be approached through digestible, bite-sized guidance—such as tooltips or an always-on AI assistant to navigate resources—rather than a rigid step-by-step onboarding process.

Iterative Mindset: Working with engineers on a complex, industry-wide challenge reinforced the importance of iterative collaboration to deliver an MVP with core features, while saving "nice-to-have" features for future iterations.

Thanks for reading!

Every project tells my different quirks—would you like to see what’s next?

Danone

SchoolCraft