The trigger is recognized
The request does not match the written lane, or something changes mid-task that the routine did not anticipate.
Useful input: a clear reason the request stopped matching the routine.
Give every exception a safe stop, not an improvised guess.
When a request does not match the written lane, the right move is not silence and it is not quiet improvisation. It is a short, repeatable procedure that protects the client and keeps the decision with its proper owner.
Follow one exception from trigger to resumption.
Work stops at a safe point
No partial change is left in an unclear state. Access, data, and service are left in a condition that can be explained.
Possible output: a paused request with a known, describable state.
Context is recorded
What was observed, what was attempted, and what is now needed go into the request record's exception fields.
Recorded result: enough detail for the decision owner to act without re-investigating.
The named owner decides
The escalation matrix identifies who decides and what information they need to do it.
Possible output: an approval, a decline, or a request for more information.
Service resumes or redirects
The original request continues, is replaced by an approved variant, or is formally moved out of the recurring lane.
Review signal: a repeated pattern here may mean the lane definition needs to change.
Safe stopping points are lane-specific, not universal.
- User supportLeave the user informed — the account or device is in a working or clearly explained state, not mid-task without an update.
- Microsoft 365Leave the tenant consistent — a partially applied license, access, or configuration change is completed to a known-good state or fully reverted, never left ambiguous.
- Managed cadenceLeave the record honest — an interrupted baseline or endpoint routine is marked incomplete rather than reported as done.
These examples do not imply a particular tool or automation; they describe the standard the stopping point should meet.
A representative exception note, not a customer record.
Observed: the requested change would affect a shared mailbox outside the documented routine.
Attempted: none — the change was not applied.
Needed: confirmation that the shared mailbox is in scope, and who can approve the change.
The full field structure lives in the delivery pack's exception-note template.
Ask these when a procedure is only followed sometimes.
- Does everyone stop at the same point?Or does it depend on who happens to be working the request?
- Is context recorded before the decision comes back, or after?Recording it afterward usually means details are missing.
- Does the client receive an honest update?Silence while an exception is open erodes trust faster than a plain explanation.
- Does a resolved exception change the written lane?Or only the one ticket, leaving the next occurrence to repeat the same detour.
Exceptions are useful data for the next review.
A single exception is a decision. A pattern of them is a signal the lane definition itself needs attention.