Crazy it's taken them so long to even think of this, never mind implement it.
I believe that there used to be a website that did this for you, that Ancestry requested to cease and desist. Yet they've taken years to actually roll this out.
I also have a script I built to download all matches and then build auto-clusters. It took about three days, with almost all of the work on figuring out on how to best cluster matches. Don't know why it takes them a year to just roll out 64 groups (which is still not enough).
There are so many tools Ancestry could make, but it seems they are very slow. Another I use is compare matches between two tests. This is useful as Ancestry only show some common matches - not sure exactly how that works as they obviously don't all share the same MRCAs. So with something like a 3rd cousin, you might share a few hundred matches, while Ancestry only shows maybe 20-30 common matches. When those 'hidden' shared matches form themselves into clusters, while others are just coincidental via non-shared ancestry. When you have mutual clusters, that Ancestry doesn't show as shared, the most likely explanation is shared ancestry, while not sharing the same 8cM+ cluster as yourself.