Ask HN: What are you doing to deploy your LLMs with Enterprises?

Given the amount of resources (both $$ and time) to train an LLM, how are you doing private deployments of these costly models for enterprises? Once deployed in private cloud / on-prem, how do you gather any usage analytics, enforce license or prevent model misuse?