How it came together
Team leads at fast-moving startups, managing anywhere from 5 to 15 people across several projects, expected to catch problems before they turn into fire drills.
Status lived everywhere except the dashboard, scattered across standups and Slack messages. Every tool could tell you what had already gone wrong. None of them could tell you what was about to.
A missed risk signal meant a late launch, a quietly burnt-out teammate, or a lead spending their week chasing updates instead of unblocking anyone.
I was given two days, full creative freedom, and one hard requirement: merge familiar dashboard UI with a real AI feature, not a chatbot bolted onto the side.
I mapped the user and their pain points first, then designed the brief's Home and Analytics screens, including an AI Daily Digest that turned scattered status into one clear read every morning.
The brief was done. But finishing an assignment isn't the same as finishing the thinking, so I kept asking questions. The Digest was fast to read, sure, but was it fast to act on? Someone still had to notice the problem, decide what to do, and go make the change by hand.
So I designed one step further than what was asked: an AI Recommended Action card that closes that gap. It names the exact problem, proposes one specific fix with the numbers already worked out, and asks before acting. There's a genuine "do something else" option too, so it's never forced on the user.
Between the brief and the iteration, the dashboard went from reporting status to actually reducing the work of acting on it. Every card and transition is fully prototyped, with hover, press, and animated states, so it reads like a working product instead of a static comp.
Finishing the brief is easy to mistake for finishing the thinking. The habit worth keeping is asking "what's still causing friction?" even after the assignment is technically done. That's where this feature came from.