This week - we're getting clarity on agentic AI for the enterprise, but what should be built - and how? LLMs require different security measures, and active inference AI shakes up the conversation.