AI Companion for Insurance Policyholders
Need is an AI health tech company building a cancer protection system that enables anyone, anywhere, to achieve the best cancer outcome. As a lead product designer, I redesigned the end-to-end Need mobile experience for cancer prevention and early detection in a one-month initiative, driving a 15% increase in MAU.
CHALLENGE & OPPORTUNITY
Need 1.0 had low engagement at 4% MAU, and no insurance app was AI-first
Need is distributed through insurers, so engagement drives risk and claims, and App 1.0 was stuck at 4% MAU. In a category where most apps are one-way and transactional, an AI-first approach could guide users, answer health questions, and drive engagement beyond insurance itself.
USERS
Cancer insurance policyholders navigating cancer concerns or treatment
Most cancer insurance policyholders we serve are over 55, often showing up at moments of high emotional and cognitive load. Treatment itself can bring fatigue, vision changes, and cognitive fog.
SOLUTION 1
Clear starting point with accessibility
To support older users, the home screen leads with recognition over recall: suggestion chips for common questions, action shortcuts for high-stakes tasks, and a persistent mic so voice is first-class. Type and touch targets meet WCAG 2.5.5.
SOLUTION 2
Provide instant support and get users connected with Need nurses
Users can ask questions or report cancer-related symptoms. When needed, cases are escalated to Need's dedicated nurses. This experience boosts credibility and addresses initial concerns about AI's lack of warmth and trustworthiness.
VALIDATION
Refined the conversation flow and LLM tone & voice based on user testing
One of the biggest insights from the user testing session was that the responses felt too robotic, preventing users from engaging more. To address this, I developed the LLM Tone & Voice guidelines, refined prompts, and evaluated responses for both accuracy and tone.
SOLUTION · LOCALIZATION
Carry the same accessibility and voice into Korean and Japanese
Need launched in Korea and Japan, so accessibility and voice had to hold in both. Type metrics were re-tuned per script (CJK needs different baselines than Latin), and prompts and tone were written natively per language, not translated, so the AI reads as locally fluent.
PRODUCT MARKETING
Achieved an 89% completion rate of the interactive demo
Following a successful launch and strengthened confidence from our business partner, we had the opportunity to collaborate on a marketing initiative that allowed potential customers to try the app. It achieved a high completion rate of 89%, validating strong interest and demand from policyholders.
