AI's Binding Boxes

We're all used to seeing these images now that have been labeled with bounding boxes around objects. But does that box have any other meaning to the AI in terms of what it has found, or is it simply a convenience mechanism and affordance for our eyes? This is how AI perceives, it says within this area, there is this object. And it's the models are very, let's say, computationally more cost effective to do that because it's just a convenience part. If you really want to do very edge-based semantic segmentation type of tasks, those costs can be much higher.