PwC found that the companies getting the best financial results from AI are increasing the number of decisions made without human intervention at almost three times the rate of everyone else. That's not a future scenario, it's already happening inside finance functions right now, and most governance frameworks were written for a world where a human always made the final call. This keynote looks at what changes when AI stops assisting decisions and starts making them: where controls need to move from reviewing outputs to constraining behavior, how audit and SOX frameworks hold up when there's no single decision-maker to interview, and what accountability actually means when a model, not a person, chose the outcome. Most AI governance conversations stop at responsible-AI principles and ethics statements. This one is about the mechanics: what specifically breaks in existing control frameworks once AI is making autonomous calls, and what CFOs are actually building to replace it. Key takeaways: - A working definition of where "autonomous" starts in your own finance function, for governance purposes - Three places most SOX and audit frameworks assume a human decision-maker that no longer exists - What to ask your AI vendors about decision logs and reversibility before scaling autonomy further