AI build guide
Build a dashboard with AI that answers real questions
A dashboard succeeds when a person can make a decision confidently. Begin with questions and metric definitions, not a collection of charts.
01
Define every metric
Specify source, grain, filters, time zone, and edge cases for each number.
02
Design for comparison
Use hierarchy, units, baselines, and accessible alternatives so changes are understandable.
03
Validate data and speed
Test empty, delayed, partial, and conflicting data as well as the normal path.
04
A production-shaped Xroga prompt
Build an operations dashboard for weekly activation, retention, and support volume. Define every metric, source, time zone, and filter; show freshness and missing-data states; use accessible tables alongside charts; and test slow and partial responses.
05
Evidence to require
Trace displayed numbers to a query or fixture, test boundary dates, and verify keyboard and mobile access. Attractive charts are not evidence that the metric is correct.
06
Before calling it production-ready
Review permissions, data handling, error states, accessibility, mobile behaviour, tests, environment configuration, observability, and rollback. The exact checklist must follow the project rather than a fixed framework template.