Live: pocketseed.io

The team arrived certain and over-built. I found what the certainty stood on, and shipped only that.

Role: Sole designer Year: 2026 Sector: Claims management Delivered: User interviews and synthesis · refined v1 scope · working AI-built prototype · reusable design system · engineering handover Outcome: Shipped. Live at pocketseed.io

Their MVP was sized for a vision rather than a customer. I ran the interviews that tested it, and three things surfaced: there was no shared need across the segments they were targeting, trust in AI was low, and the real objections were security and workflow rather than capability.

That reset the scope. I cut the v1 to what the first clients actually needed, built the prototype in AI tooling so the engineers received something working instead of a spec to interpret, and handed over a reusable design system with it.

The interesting part wasn't that they'd built too much. It's that their certainty was real, just attached to the wrong thing. Testing a vision isn't asking whether people like it; it's finding which single claim underneath it is load-bearing, and letting the rest go.