What had to change
Enterprise AI and cloud consoles fail when complex configuration, permissions, async state, and error recovery are hidden behind pretty dashboards.
Why this version is sanitised
The public artifact is intentionally sanitized: it shows the workflow shape, state model, and UX decisions without exposing employer data, customer screens, internal URLs, or private incidents.
The constraints that decided it
- High-density configuration had to stay scannable for repeated operations work.
- Async tasks needed visible status, retry, failure, and handoff states.
- AI gateway actions needed permission boundaries, review points, and traceable logs.
- Support and operations users needed enough detail to debug without jumping between five tools.
How it actually runs
- 01
User changes a cloud service or AI gateway route and sees validation before submit.
- 02
The submit path creates a task with owner, status, diff, rollout scope, and rollback affordance.
- 03
Logs, metrics, and task history are attached to the same operational object.
- 04
Failures are turned into actionable states instead of opaque toast errors.
Surfaces
route configuration · task queue · log search · change review · failure recovery