Gmail's important inbox requires training to be useful. It also has a very high false positive rate (lots marked as important that really isn't). We use a comb

Gmail's important inbox requires training to be useful. It also has a very high false positive rate (lots marked as important that really isn't). We use a combination of algorithms to make a smarter automated split, but can still learn if the user recategorizes an email. It's not buzzwords - it's science!