The final routing decision for local LLM deployment in a network lab. Matching specific AI models to voice loops, background tasks, and critical operations.