# Cost and latency dashboard

Use this dashboard outline to keep AI runtime economics visible as usage grows.

## Dashboard panels

| Panel | Signal |
| --- | --- |
| Provider mix | Share of traffic by model, provider, region, and fallback |
| Cost per task | Cost by workflow, user segment, tenant, and release |
| Latency budget | P50, P95, timeout, retry, and queue time |
| Quality correlation | Eval score, reviewer override, and user feedback by route |
| Drift alerts | Spend, latency, fallback, or quality movement outside bounds |

## Review cadence

- Daily during launches and incidents.
- Weekly for operating review and route tuning.
- Monthly for procurement, architecture, and roadmap decisions.

## Output

A runtime view that connects cost and latency to user-visible quality and operational choices.