My initial instinct was to introduce a dedicated modal, but I hit strong pushback to reuse the existing table and avoid creating a separate experience.
To make the tradeoff concrete, I designed the table-based approach in full fidelity. Seeing new uploads intermingled with approved evidence made the issue clear: it was harder to scan, harder to understand what was pending, and easier to miss actions.
I partnered with engineering to assess effort, and both approaches were estimated to take roughly the same time. With implementation cost no longer a differentiator, we aligned on the modal. It simplified the reviewer experience, made states more explicit, and reduced the risk of long-term complexity and edge cases.

Other lofi exploration: using existing table to display everything in on page