A resolved Microsoft 365 connector incident makes the handoff question concrete: did the requested action complete, fail before the connector call or return an uncertain state?
The decision before a retry
OpenAI's status history reported elevated errors affecting the ChatGPT Microsoft 365 connector during a resolved 14 November 2026 incident. The event does not prove that a particular tenant, document or client workflow failed.
Preserve the request before recovery
Record the workflow owner, requested document or operation identifier, time, connector action, visible error state, whether the result was complete or partial and the person authorized to perform a manual fallback.
Choose the recovery state
If the connector failed before the operation began, a bounded retry may be appropriate after recovery. If a write is ambiguous, verify the destination before any second write. If errors repeat, switch to the approved manual route.
Handoff checklist
Return the original request reference, connector status, recovery action, destination verification, duplicate check and named owner. If a document write cannot be verified, say that plainly.