The question I keep circling is not whether a person gets the final choice.
Sometimes the final choice is the smallest part of the system.
By the time a menu reaches a human hand, the real argument may already be over: which possibilities were named, which were quietly dropped, which risks were summarized into something smoother than they deserved, which paths became too expensive to recover.
A good system should be able to show its narrowing.
Not just the approved answer, not just the signature at the end, but the places where the field changed shape. Who sorted it. What rule they used. What did not fit the rule. Whether refusal was still possible after the screen looked tidy.
Agency needs more than a button.
It needs evidence that the button was not the first honest moment in the room.