Exploration began with two gradient directions: saturated and subdued. Seeing them in-situ made the saturation problem immediately apparent. When multiple gen AI features appeared in a single experience, a rich gradient on every entry point created visual noise that undercut the hierarchy we were trying to establish. We pivoted to a more limited approach early. From there, I worked through color combinations with attention to hue selection, gradient stop locations, and blending behavior. The goal was a gradient that felt harmonious and not muddy. We landed on a red, magenta, and indigo combination, close enough to previously used product gradients to reduce user friction and visually distinct enough to read as a system-level signal. The gradient is angled within the button component, maintaining visual fidelity as the button is responsively resized. I tested several versions of the implementation guidelines with product teams and learned that the logic needed to be simple enough for any team to apply correctly on their own. The final guidance was precise, and I consulted individual product teams as-needed. These constraints kept individual product experiences from oversaturating while preserving the system's legibility across the suite.