Financial institutions aren’t short of intelligence. They are short of agents they will let act. Running AI agents in regulated finance requires bounded permissions, traceable actions and defensible outcomes. The challenge is no longer just what AI can do, but how institutions can govern what it does and prove it.