The segmentation model Adobe uses is competent. On human subjects with clean backgrounds, the cutout is clean and the edges hold up at normal viewing sizes. On product photos against solid surfaces, results are reliable.
Where it struggles is the same place most bundled tools struggle: fine hair, fur, and semi-transparent materials. Adobe's model produces a binary mask in most cases, meaning a pixel is either fully opaque or fully transparent. There is no partial alpha for the wispy strands around a person's head.
For a portrait to use as a LinkedIn photo, that is fine. For a portrait to composite over a coloured background where the hair edge will be visible, you will notice the difference.