The other form of automation is the more symbolic systems approach where each step leads to different automated outcomes. This version is more like a maze – there are a lot of options, but in the end only a finite amount of directions you can take (you cannot, for instance, fly up on the z-axis out of the maze, that would be cheating).
This paper examines one example of the earlier non-machine-learning approaches to to automation in the field of automated phone calls. What's good? What's bad? And, what's next?