Should data augmentation be encouraged?

As we are talking about data centric, should we artificially “create” data or rather focus on what we have and optimize what we have?