The cited guidance identifies two ongoing maintenance activities for a connected add-on: checking connections for warnings or errors and monitoring the functionality and behavior of deployed AI components in production. A connection showing a warning or error icon is handled by selecting it and choosing Fix connection or Re-authenticate; deployed components are monitored while operating in production.
How to check it
| Area | What is checked | Supported action |
|---|---|---|
| Connection health | Warning or error icons on connections | The flagged connection is selected, then Fix connection or Re-authenticate is chosen. |
| Production behavior | The functionality and behavior of the AI system and components identified in the map function | Those components are monitored while deployed in production. |
The connection-maintenance action is conditional on a visible warning or error. The production-monitoring expectation applies while the components are in production. The cited statements do not establish a routine review frequency for either activity.
What a team must still confirm
The cited statements do not specify:
- who owns maintenance, coverage hours, or escalation responsibilities;
- how often checks should occur or what thresholds should trigger action;
- which components fall within the referenced map function;
- what records, test evidence, or incident information should be retained;
- which change-control, security, privacy, backup, or recovery procedures apply;
- whether support fees, contract terms, service levels, or other commitments apply.
The absence of these details from the cited statements does not mean they are unnecessary. It does mean that a complete maintenance policy cannot be inferred, and neither maintenance action should be presented as guaranteeing availability or performance.