I'm writing a book chapter at the moment on system design, and I realized that there is only one firm recommendation that I can make about when a system should implement its brains in digital-land as opposed to analog- land (or, for that matter, mechanical-land).
The guiding light? Overall system cost. How to get there? The collected wisdom of the team.
Sometimes that even means that while the _absolute optimum_ given the technology available on the day you decide may be to go down a particular path, the best bet _on that day_, _with that team_ may be different, because with the people that _you_ have available _right then_, the "we can do it and understand it" may trump "this is universally best".
Of course, sometimes leaning too heavily on the team's current skill set means that you're either letting your company's designs get stale, or that you're building products that try to solve problems with a golden hammer that are better solved with a rock and a string.