I think AI could be of use for Poser in Poses. Poses are vectorized information, and there is plenty of it. It would only need a qualification that could link it to terms in the prompt. Problem, as ever with AI, is ownership of data.

As far as image enhancement is concerned I think it is better if Poser stays true to it's roots: provide template imagery/geometry of posed/animated figures and leave the final workout to other apps dedicated to that task. There is no way a small company like Bondware could ever stay par with development in the sector. 

In another thread on this subject I showed the below AI enhanced image:

Guess the Poser figures used.