Depending on what you mean by AI exactly, any language is good, as there are matrix/vector operations underneath any AI algorithm and every language has a capacity to do that.
However, if you mean AI in a popular sense, you should focus on a language that allows you to develop a closed-end application.
As for the job part -- anything from data exploration to algorithm implementation and app development can be a part of such a job.
LISP and its dialects can read their code, include it as string, alter it and return it to the code.
This feature looks very promising to me since it allows a dynamic codebase that optimizes itself depending on the task. At least in theory.