Dashboards are a commodity. If your users can read your charts correctly and still can't name a next step, more visualization won't fix it — the information model will.

A two-week engagement to redesign your data or AI-output surface around decisions instead of display. I did exactly this at Veltrix AI (week-4 retention 22% → 49%), Luna Health AI (correct self-reading 45% → 82%) and Dreamteam (weekly return 18% → 41%).

Answer first, evidence on demand. Lead with the finding in plain language; the chart sits underneath as proof, not as the headline.

Make the AI legible. Every output traceable to its numbers and source. Low-confidence results framed as questions, not claims. In anything financial or clinical, trust is the product.

One decision at a time. A single prioritized action per insight, alternatives tucked behind. Ending paralysis, not relocating it.

One atomic unit, every surface. Design a single insight/metric component well and it scales from marketing site to product to mobile without redesigning trust each time.

Information hierarchy and dashboard IA · chart and gauge selection (and what to delete) · empty, loading, partial-data and error states · confidence and uncertainty treatment · responsive and mobile behavior · a reusable data-viz component set in Figma.