Modernised inside the codebase it already had

The product runs on Java FX and Swing, and an acquired codebase sets hard limits on what a redesign is even allowed to propose. Designing first and discovering the constraint later wastes everyone’s time, so the early part of this project was spent working out what the framework would and would not do.

AI was useful here as a way to get up to speed on an unfamiliar stack fast — enough to ask engineering better questions, sooner. It did not replace that conversation, and we later built a proper process for it with the engineers. What it shortened was the stretch where I did not yet know enough to ask anything useful.

A redesign is easy to dismiss when it sounds like a designer's opinion, so I did not lead with one. I brought churn data. It showed the outdated experience was driving users away, and that the fix was a full pivot rather than another round of incremental cleanup. Leadership changed its position and backed the redesign.

The obvious follow-on to "full pivot" is "full re-platform," which is expensive and slow. I designed the modernization to fit the legacy Java constraints instead, and worked closely with engineering on creative workarounds wherever the framework would not do what the design asked for. The constraint set the design.