Case Study
2025
Agentic AI in Health Insurance
Designing a proof of concept agentic AI assistant to streamline health insurance processing. Every day, thousands of health insurance applications move through a maze of manual checks, policy rules, and human judgment. It’s slow, repetitive, and error-prone — but every decision matters to someone waiting for coverage.
Our goal was to help — not replace — human processors. We explored how could AI and people work side by side to make complex decisions faster, clearer, and more humane.
Industry
Healthcare
Role
Lead User Experience
Company
Publicis Sapient
Challenge
Health insurance processing is notoriously slow, manual, and prone to human error. Reviewers spend hours cross-checking applications, verifying eligibility, and interpreting policy rules — a process that’s as high-stakes as it is repetitive.
The organization wanted to explore how agentic AI could help automate these tasks without compromising accuracy or compliance. But the real challenge wasn’t just technical — it was human:
How do you introduce AI into a regulated, trust-sensitive workflow?
How do you keep users in control while still gaining efficiency?
And how do you make complex AI reasoning transparent enough for non-technical reviewers to trust?
These questions shaped every design decision, from the workflow architecture to the smallest microinteraction. The goal was clear: create an AI assistant that speeds up work — and earns user confidence with every interaction.
Process
01
Strategy
Worked cross-functionally with strategists, development, and product experts.
02
Visual Design
Created clear and well defined look and feel with strong visual hierarchy.
03
Prototyping
I built an interactive prototype showing how the agent would flag, explain, and resolve issues across an application. Each action included a clear “Review → Approve → Override” loop to maintain user agency.
Conclusion
This project explored how agentic AI can simplify complex insurance workflows while preserving human oversight and trust.
Prototype
Simple flow
User starts out by seeing applications submitted for coverage, selects an application the was processed by A.I. and needs human review. The processor can review the A.I. process and approve or deny the recommendation.