The pains
Cost reports do not result in shipped fixes
You can see the waste, but routing the fix through tickets, owners, IaC and CI eats the savings before they land.
Manual remediation is risky
Engineers fear shutting down a "stopped" VM because last quarter it turned out to be load-bearing for a forgotten dev workflow.
Kubernetes cost is opaque
Node-pool spend and pod requests drift quietly; nobody owns "the cluster bill" until the EKS / GKE invoice doubles.