My threshold for considering automation is doing it once :) but if the work to automate is non-trivial, then I usually don't get around to it until the 3rd cycle.
I usually evaluate a few things, in this order:
1. How much time will it take to do manually vs automated?
2. What are the chances I'm going to have to do this again?
3. Is it useful to just me, or other people too?
4. Do I get any value by doing this without automation? (e.g. will I gain insight by taking time to go through this process, that I would miss if I built a tool to just do it)
...but I suppose that's mostly for "tasks" whereas you mean "processes" in the sense that a process is more difficult to automate because it has more steps and dependencies. In that case, I think there are more evaluation criteria like:
-What is the risk of doing this manually and screwing it up?
