This is a concept build. There's no client and no production data behind it. It shows how we approach dashboards that people actually trust.
The problem
Most dashboards answer "what" and leave people guessing at "why". Numbers move, nobody knows whether the change is real, seasonal or a data problem, and the dashboard slowly becomes a screen people glance at and ignore.
What we built
A dashboard with the usual charts and metrics, plus an assistant sitting beside them. Ask it why a number changed and it explains in plain language, pointing at the data it used, so the answer can be checked rather than taken on faith.
- Metric tiles and trend charts for the handful of numbers that matter
- Anomaly highlighting, so unusual movement is flagged rather than buried
- An explaining assistant that says what changed, why it probably changed, and where that came from
How we'd approach it for real
We'd start with the decisions the dashboard is meant to support, not the data you happen to have. Then the metric definitions, agreed and written down. The assistant comes last, grounded in those definitions and the underlying data, and evaluated before anyone relies on it.
What this demonstrates
AI where it adds value: here that means explaining, not replacing, the numbers. And provenance by default, so every answer is traceable to its source.