Automation won’t “save” you completely, but done well it can reduce the number of points of human failure. Human failure is by far the largest source of systems failure, but the job of anyone working with systems is not to remove or punish humans who fail, but to make it harder for them to fail in the first place. One excellent way to do this is to reward failure, as that person has surfaced a point where failure is possible so that it can be mitigated or removed in the future. In your example, you could mitigate the chance for future failure by not giving production access to standard dev accounts. Require devs/admins use a different account when they need to do stuff in production, and give that account access to _only_ production. You could go even further by designing a system that requires secondary authorization to do anything to production.