We get asked to evaluate a Kubernetes migration a few times a year, and the honest answer is usually "not yet." Not because Kubernetes is bad — it solves real problems — but because most of the teams asking don't have the problems it solves.
The signal we actually look for: do you have enough distinct services, deployed independently enough, that the orchestration overhead pays for itself? A single monolith or a handful of services rarely clears that bar. A few dozen services with independent scaling needs usually does.
The cost side is real too: someone on the team needs to actually understand the failure modes, not just the happy path. We've seen more outages caused by an under-understood Kubernetes setup than by the plain VM setup it replaced. If nobody on the team wants to own that expertise, that's a legitimate reason to wait, not a gap to paper over with a managed service.