General explainability in AI is very difficult — I believe that at this stage in the development of AI it is most effective to tackle explainability in context of specific domains/applications. Pranav Kuber and the team @ Piction Health have a great implementation of domain specific explainability - blog post here.
