Automated data normalization

I regularly have to take in large volumes of data that are not normalized, and then spend time reducing their size and storage requirements by simplifying them. What I would like to see is some intelligent functionality that can analyse a table to identify columns that should be normalized, and then when selected can be split out to a separate reference table with a reference key which can be added to the source table. As this needs to handle data refreshes, it should be encapsulated in the M language steps for data loading. It should be simple enough for non-experts to use.