Why it stopped and asked me
Destructive and high-risk calls wait for a person.
On this page
An agent pauses when a call is destructive, high risk, or expected to touch more records than the bulk threshold. It records exactly what it wanted to do and waits.
Your decision covers one call
Approving does not grant an agent permission to delete customers. It grants permission to delete customer 100, once, with the values you were shown.
The approval stores a fingerprint of those exact arguments, recomputed when the call actually runs. If anything about it changed in between, the approval no longer applies and the call is gated again. That is why the approval screen shows you the raw values rather than a summary — the values *are* the decision.
After you decide
Approving continues the run from where it stopped. Rejecting tells the agent it was declined — with your reason — so it can respond sensibly rather than pretending it succeeded.
A request that nobody answers expires, and the run is stopped. An agent's plan goes stale, and approving a day-old deletion is not the decision you think you are making.
Who can decide
Owners, admins and managers. Viewing the queue and deciding on it are separate permissions on purpose: reviewing and authorising are different jobs.