Designer-led, AI-built:
Benefits enrollment experience
I designed and built an interactive, state-aware benefits enrollment platform using Lovable, with Claude as a coding partner.
Every product decision — structure, flows, state logic — was mine. AI handled implementation. This project explores what changes when execution is no longer the constraint, and where product thinking becomes the differentiator.
The problem
Benefits enrollment is high-stakes, low-frequency, and confusing. Most platforms treat it like a form to complete rather than a decision to support.
Two user needs shaped the system:
Deliberate enrollees need guidance, context, and time
Power users want to compare options and choose quickly
More than just UI, this a product architecture problem.
Finding the product
I explored three homepage directions:
v1: Dashboard — familiar, but indistinguishable from existing platforms
v2: Enrollment flow — added progress + context, but duplicated core information
v3: Editorial context layer — removed selection entirely
Decision:
Homepage = why and when (context, deadlines, guidance)
Plans = what and how (selection, comparison, actions)
Each page has one job.
Key product decisions
1. State-aware system
The product adapts based on where the user is in enrollment.
FSA/HSA is disabled until a health plan is selected
Advisor messaging evolves from general → specific
Notifications reflect real progress
No UI references actions the user hasn’t taken yet
Complexity appears only when it becomes relevant.
2. Guided + self-serve paths
The system supports two modes without forcing either:
Guided: short assessment → recommendation with reasoning
Self-serve: skip directly to plan comparison
If the recommendation doesn’t change based on inputs, the flow is removed.
3. Cost as a running decision
Cost isn’t shown at the end — it updates as users choose.
Running cost bar reflects current selections
Plan details show projected annual costs (not just premiums)
Supports real tradeoff thinking, not surface-level comparison
What AI changed
Multi-page prototype in days
Rapid iteration across product directions
What it didn’t
Product decisions
System logic
Accessibility correctness
Pattern judgment
The outcome
A working, state-aware enrollment system that:
Separates understanding from decision-making
Adapts to user progress in real time
Supports both guided and self-serve behavior