If the AI opens the comments, who sets the conversation?

At the Financial Times, the AI work described publicly sits at the entrance to the comment thread: Utopia supports moderation, and a detailed prompt can generate a "Join the Conversation" question in seconds. The community team reviews, edits, and publishes it. 1

INMA reports that more than half of FT subscribers read comments, while only about 4% write them. It also reports higher engagement among comment writers and readers, plus higher retention among subscribers who comment. Those figures are reported associations, not proof that the AI caused the outcome. 1 Nieman Lab's independent account makes the boundary plain: automation can reduce moderation noise, but context, policy, and human judgment still shape a healthy discussion. 2

That leaves the Agentic Media question: if a system helps decide the first prompt, is it merely moderation, or part of editorial framing? Where would you draw the line: filtering abuse, inviting replies, or framing the story?