Connected Care

I was working on multiple AI projects for Microsoft Research, including building proof of concepts with GPT-3 from OpenAI. One of them was Windcrest in Microsoft Health Futures, which later became Connected Care.

Our mission was to prove customer interest, design and innovate with AI, and find the product concepts a home in the Health and Life Sciences team to inform their work for patient engagement.

The aging population is one of the greatest social, economic, and political transformations of our time.

Caregivers are a diverse and global population. 1 in 5 adults is a family caregiver. Consider that almost 20% of the Microsoft population potentially.

We took a 3-wave product strategy of research, proving customer interest, innovation, and then product fit to achieve patient and caregiver engagement in the home — what we called the "Houspital."

We completed these 3 waves in 18 months.

We used the Kano model to help prioritize what to build by categorizing features as Must-Be, Performance, Attractive (delighters), or Indifferent.

The rule of thumb: prioritize Must-Be first, followed by Performance, then Attractive. Because we were building something pre-launch, anything categorized as Performance was treated as similar to Must-Be.

I assigned designers to pillar areas and guided them on applying brand and information architecture to the 3 pillars.

The process included an infinite customer feedback loop with an internal caregiver council. I established the care council and empowered designers to meet with the council regularly to get feedback on designs.

I identified each workstream of the design team and worked with each designer to generate our payload roadmap to get us to an Alpha launch.

We aligned on a CRUD model and architecture, including principals around the structure to easily make decisions without meeting regularly. This gave the engineering team the building blocks to get started.

To accelerate the engineers, I had the team start with Microsoft Fluent — but it wasn't "warm" enough for caregivers. So I empowered my principal visual designer to theme it.

I led the entire design team to take the Fluent system and apply a theme using the brand elements, then coached the team to apply the brand to the CRUD framework.

We used the socio-ecological model to understand caregiver needs across individual, interpersonal, organizational, and community layers.

I took the learnings, partnering with the researcher, and mapped a customer appointment journey with the potential LLM capability impact. Then I led a workshop with the entire team to design for this journey.

I created a fully-functional, high-fidelity prototype of the new flows using Figma. We recruited subjects who fit our criteria and ran 6 usability tests in the first round and 3 after iterating on the issues we identified.

Issue 01 — Distinction between AI and user-generated content: Users wanted a very clear understanding of where the content came from, whether it was AI-generated or user-generated.

We provided visualizations of key information, added privacy controls, and defined a clear purpose for each section.

Our vision aligns with the KLAS patient engagement platform and elements that define the best places for aging individuals — security, health, personal capabilities, and an environment that supports them.

We address these crucial aspects by delivering health-centric services directly into people's homes. This is achieved by engaging users effectively — not in just one part of the journey but the entire one — establishing trust with health consumers and offering a cohesive care journey that spans between users and providers.