I am thinking that it might be possible to use AI to scan a previously rendered version of a file, and as it's rendering a new version/revision, if any of the frames from the new render are identical to a previous version, the AI would pull those identical frames from a previous version so only the new frames would have to be rendered when producing a new version. This would make it easier and faster to make revisions theoretically.