Design exercise: making AI-surfaced research insights something a team can trust and act on.

Problem: Research insights get lost across disconnected tools, and AI can spot patterns in them, but only if people can trace where each one came from.

Approach: One core flow - Discover, Compare, Verify. Scan AI-surfaced insights at a glance, see how evidence clusters into themes, then trace any single claim back to its exact source before a human signs off.

Key decisions: Confidence scores are a relative signal, not ground truth, always paired with a traceable source. Violet marks anything AI-touched, everywhere. Plain language and calm typography keep a data-dense tool feeling human, not clinical.