Most of my Prolog experience was in academics - for my own research as well as for maintaining some existing packages. But I did get to use it outside as well. There are many cases where a controlled language is needed or comes in handy and solves a lot of issues. This is where Prolog shines - plus you get to use a lot of the previous stuff you developed as well.
When it comes to an "AI" language vs. typical language with an "AI" library, I will just repeat a phrase that is often used in the field: what you thinnk of as an AI problem today will no longer be considered AI tomorrow, since it will then be well-understood and implemented in many places. People used to say that in response to the (seeming lack of) progress AI had made to date. A good example of that is Siri or Amazon Echo. Not many people consider it an AI problem anymore, or so it seems.
When it comes to an "AI" language vs. typical language with an "AI" library, I will just repeat a phrase that is often used in the field: what you thinnk of as an AI problem today will no longer be considered AI tomorrow, since it will then be well-understood and implemented in many places. People used to say that in response to the (seeming lack of) progress AI had made to date. A good example of that is Siri or Amazon Echo. Not many people consider it an AI problem anymore, or so it seems.