In any sizable organization, you can't get around automatic updates.

But updates should be rolled out slowly and you need enough telemetry to detect problems as it's rolled out. Reboots, crashes, cpu/memory use, end user reports, etc should all be used to detect issues and pause the rollout.