This is a great distinction between an AI that simply provides information and one that can actually help resolve the issue. The Actions and Procedures side is especially important because real support often involves multiple steps, system lookups, and decisions not just finding an answer in a knowledge base. I also agree that the quality of the integrations and guardrails may matter more than the AI itself. If the agent can safely access the right systems and knows when to escalate, it can take a meaningful amount of work away from human support teams without trying to replace them entirely.