"You can't proactively optimize your infrastructure without understanding the business running on it".
When knowing the business, you can create many dedicated KPIs to monitor your optimization factor proactively, and instead of relying on "resource" data, you can rely on "business" data.
What would that look like?
Kube pods don't only take resources from an instance - they should serve a purpose. Know what the business that is running on them? Use it as a KPI - number of running threads, number of requests per minute, etc. Take the cost of the infrastructure, attach to it the business KPI you can measure, and then you can understand if the pod is actually optimized or not.
S3 bucket optimizations - when you know the business data you store in a bucket, you can understand proactively how the usage and access patterns would be. Saving logs for example would be useful for ~3 days, and then move to colder storage or terminate. So, instead of "fixing" this after provisioning, it'll take a few simple questions to already apply the right lifecycle policies on the bucket even before pushing objects into it.
Simple, right?