It took me barely five minutes to convince my boss we shouldn't move everything to the cloud — a quick estimate on the AWS calculator was enough.
But he's still keen to try Kubernetes. I think we can improve our deployment process (currently custom scripts) and introduce genuine load balancing, but there's a big increase in complexity. The custom scripts aren't very long, I suspect we'll need at least the same amount of configuration for Kubernetes, plus we add Kubernetes itself.
He's hoping for more ops understanding/interest from developers, and which is probably reasonable, but at the cost of less understanding of the whole system from the ops staff. I think there's some CV-driven-development too.
Step 1 will have to be us documenting what we expect to achieve from Kubernetes, and how we will measure achievement.