It should provide all the same benefits provided by the icon that shows that an ID was picked from the list generated by the computer vision thingy, namely making it much easier to tell whether an identification should be double-checked or not. Note that the computer vision icon also comes with no information about the motivation behind its use.
To a certain extent, that is a desired outcome of this change. However, there are several different places where there are “Agree?” buttons. The worst one is on the identification thumbnails page, where you can’t see if an observation has more photos, or comments, or see the location and date of the observation before you click the button. I suspect you are using one of the other ones. I would vote for removing the worst one if that topic were a feature request.
I’m unfamiliar with the apps, so there may be some “Agree?” buttons there which are getting frequently misused, as well. The point is, we don’t know, so it would be good to track their use.