End-to-end product redesign for an AI-driven InsurTech platform
Challenge
Verikai's platform asked insurance underwriters to act on machine-learning risk scores they couldn't interrogate. The interface had grown organically around the model: inconsistent patterns, fragmented workflows, and predictions presented without the evidence and context a person needs to calibrate trust in them.
Approach
Led a comprehensive redesign starting from user research with underwriters and workflow mapping. Established a unified design system and data visualization patterns that made the model's output legible: what a risk score drew on, where confidence was strong or weak, and what to check before acting on it. Worked closely with engineering and data science to keep every pattern faithful to what the model actually knew.
Impact
The redesigned platform made model output something underwriters could read, question, and defend to their own stakeholders. That design foundation contributed to Verikai's positioning through its ~$120M acquisition by American Financial Group, and it is the clearest early version of the question that now runs through my work: how a system should show its evidence so people calibrate trust correctly.
Product Design
Human-AI Trust
Design Systems
