Yeah for context I mean doing slow rollouts and checking error rates on a new deployment is a given. That's not AI. But for an alert for a deployment that has

Yeah for context I mean doing slow rollouts and checking error rates on a new deployment is a given. That's not AI. But for an alert for a deployment that has been out for a while there is a bit of a decision to be made as whether to roll back, scale up etc. sometimes rolling back achieves nothing other than increasing the delta again when you roll forward and delaying stuff.