It's not a book but I really like the Elements of AI two-part course series that the University of Helskini provided (financial) support for: the two courses in the Elements of AI series consist of a pure theory course and the second course is a hands-on / applied course. Because not everyone who is seeking to learn about AI has programming skills the second course is a 'choose your own adventure' type course, with one track involving lots of programming, one 'middle of the road' option with just a touch of coding, and the other track not involving any programming whatsoever (suitable for executives, IMO).
https://www.elementsofai.com/
I've recommended this course (just the first one if you're a time-constrained executive) to C-suite colleagues in the past who wanted to become more informed about ML, DL and AI, but who didn't want a deeply technical explanation a la Andrew Ng's Coursera courses or similar content.
The Elements of AI courses touch upon the statistical underpinnings of the space (there's a unit on Bayes' Theorem), the societal implications of automated decision-making process (job creation, etc.), what tasks are "doable with AI today" and which tasks are definitely _not_ doable with A(G)I, etc.
Helping non-technical folks develop an intuition about what is possible with "AI" is crucial, I think, to having a workplace and a society that can talk realistically about the benefits and detriments of robotic data processing such as ML, DL, AI.
