Hugging Face Auto Train - POS vs NER

Using HuggingFace Auto Train tokenizer for the first time.

When I am selecting the columns to be used in my dataset I have to choose ‘words’ and ‘POS’; I am curious if this means this does not work well for NER models? Is the POS just a place holder and NER works fine; or should I just be using this for POS tasks. And I need to train a NER model manually?