A product manager who is a designer by training. I research, prototype and ship code myself, so ideas reach users quickly.

• A live view instead of a follow-up. Open, settled and rejected claims update from our own data, so HR can see where any claim stands at any time without waiting on a reply from their relationship manager.
• From a number to the claims behind it. Every number on the page filters the list below it, so "47 waiting" takes HR straight to those 47 claims and what each one is waiting on.
• The same picture in their inbox. HRs who don't log in get a weekly or monthly digest of what's open, what closed and what employees said about the experience.

• More volume, same team. Every batch that sends itself is one less in an ops queue. The team spends its time on the exceptions that need judgement, and taking on more accounts doesn't mean hiring more people.
• Faster turnaround. Automated batches go to the insurer as soon as they're ready instead of waiting for someone to pick them up, so changes to cover reach the insurer sooner.
• A channel for every insurer. Insurers with an API get batches directly. Insurers that work over email get them from an AI agent that sends the batch and reads the reply. Portal-only insurers get them from an RPA bot.

The problem

When a client signs, operations collects policy documents, maps health cards, sets up the deposit ledger and hands the account over. All of that was tracked in Excel sheets. Status and reality drifted apart, and nobody could say how long onboarding took.

What I did

• One source of truth. The Excel tracker is gone. Every client being onboarded, its stage and what's still missing now live in one board in the product, used by the onboarding and placement teams.
• Status comes from the data. Each card's stage is worked out from records already in Pazcare, like documents filed and health cards mapped, so nobody updates status by hand and the board can't drift from reality. Overdue counts show what's slipping.
• Reconcile lives in one click. Before sign-off, ops had to check the insurer's member list against ours line by line in Excel, and any gap surfaced later as a missing health card or a rejected claim. Now a single button runs the comparison inside the board and lists only the people who need action, so sign-off no longer waits on a manual audit.

• Audits mapping where the UI drifts from the system. Two UX audits of the partner dashboard showed where the UI had split apart: ad-hoc colours, raw pixel values and buttons styled page by page.
• Screens are built from shared, documented components. Type and spacing tokens, one button component with three types, and shared headers, tables and filters, with a lint rule that stops new code from adding raw values. All documented in Storybook.
• AI-built screens look like the rest of the app. Left to themselves, AI coding agents produce screens that look nothing like each other, and early on their work kept clashing with our existing visual language. Now they build from the same tokens and documented components as everyone else, so what they ship matches the product.