Put a group full of controllers on stage and ask where AI is allowed to touch the close unsupervised. You won't get one answer, or even two. Some run reconciliations through AI with no human check at all. Others won't let it near a journal entry without a second set of eyes. Nobody up there agrees, and nobody's backing down. That disagreement plays out live, in front of the people who'll have to defend whichever line they hold once an auditor comes asking. Key takeaways: - Where controllers currently draw the line on unsupervised AI, and why company size and regulation move it - What "audit-ready AI" actually means, past what a vendor puts in a sales deck - What would have to change, technically or regulatorily, before the line moves at all