I agree, the current atlas functionality is hugely inefficient. FYI there was some discussion starting here about possible ways to “initialize” atlases that don’t yet exist using current GBIF data (or more refined regional sources, like BONAP or USDA for the U.S.). We would probably want any such auto-populated atlases to be created in “inactive” status until visited and verified by a Curator. But that would still be a huge improvement.
This may all be a pipe dream though, since making something like this happen would probably be a major Staff effort. It would only be worth it if Atlases were going to have a big impact on site functionality beyond informing taxon splits.