Skip to main content

03 — HEALTHTECH POCS · APR–MAY 2025 · HEALTHCARE · GENAI

Validating GenAI for clinicians — from research to C-suite approval in 3 weeks

Role

UX Researcher · Product Strategist · Designer · GenAI Prompt Engineer

Team

1 Product Manager · 2 Engineers

Timeline

Apr – May 2025 · 3-week cycle per POC

Full details are under NDA. What I can share: the research approach, the design process, and the outcomes. The work involved designing and shipping multiple GenAI proof-of-concepts for clinical workflows at a Singapore healthtech company — validated with clinicians, then presented to C-suite with six-figure commercial value identified.

WHAT I DID

01

Led UX research with clinicians across multiple disciplines to map where GenAI could meaningfully reduce workflow overhead — without adding cognitive load to clinical decision-making.

02

Designed the POC interaction models end-to-end: framing the AI's role, defining input/output boundaries, and prototyping in Figma for clinical validation sessions.

03

Ran the prompt engineering work that made each POC functional — writing, testing, and iterating the AI instructions that shaped what the models produced and how clinicians experienced the output.

04

Facilitated structured feedback sessions with clinicians after each POC round, distilling findings into design iterations in the same 3-week cycle.

05

Presented validated POCs to C-suite and clinical leadership, translating design decisions and research findings into commercial terms.

06

Navigated NDA constraints throughout — designing with de-identified inputs only, ensuring no clinical data entered any AI workflow.

THE CHALLENGE

How might we leverage GenAI to generate meaningful productivity gain for clinicians?

Constraint

Clinical data is out of scope

All AI interactions had to operate on de-identified inputs only. Every POC was designed around this constraint from the start — not retrofitted later.

Constraint

Trust is earned in the room

Clinicians are skeptical of AI for good reason. Validation sessions had to surface real resistance and address it — not smooth it over. Early feedback shaped everything.

THE PROCESS — 3-WEEK CYCLE PER POC

01Discover

Clinician interviews and workflow mapping

Interviewed clinicians across disciplines to map pain points in their existing workflows. Identified where time was being lost to administrative and documentation work — and where GenAI could absorb that load without entering clinical decision territory.

02Develop

Rapid GenAI prototyping and internal testing

Built functional prototypes within a 3-week cycle per POC. Each prototype was prompt-engineered to match the specific clinical context — register, structure, constraints. Tested internally before clinician exposure to catch failure modes early.

03Deliver

Clinical and executive validation

Ran validation sessions with clinicians, incorporating feedback into a final iteration. Presented the validated POC to C-suite — framing findings in commercial value terms, not design terms. Both the POC and the process were evaluated.

NDA NOTE

Specific POC types, clinical workflow details, patient-adjacent data, and company name are under NDA. The process, research methodology, and validated outcome figures above are what I can share. I am happy to discuss the work further in a conversation — reach out via the links below.

OUTCOMES

6-figure

commercial value identified from the validated POC portfolio, with active implementation discussions initiated with the client

3 weeks

design-to-validation cycle per POC — from clinician interviews to C-suite presentation

Multiple POCs

validated with clinicians across disciplines and signed off by executive leadership

GenAIHealthTechUX ResearchProduct StrategyPrompt EngineeringNDA