I assume that you'll need a human being to input the data that the software would need:
Create a workflow with all the paths.
Assign the probability of success at each event. (Min., Best guess, and Max)
I recommend looking at David Snowden's work on complexity thinking, including the Cynefin model, to find out whether your domain is complex (in which patterns will be emergent and probably a machine program will not be able to deal with the complexity) or complicated (in which it might be possible to provide a machine with a certain amount of expertise). He also mentions other domains, including the chaotic, in which you will want to act quickly rather than plug numbers into a machine.
If the decision is one which is made again and again, with well-understood data - for instance, whether and how to trade based on current risk profiles - then, yes, it may be possible to create software to help make that decision. Even then, though, human intervention may occasionally be required.
Otherwise, the biggest problem with decision making is not making the decision as much as finding the information needed to make the best decision in the first place. Also, human beings are awful at handling probability, so our guesses for min, best and max are likely to include significant bias.
Unless you actually create a computer program to collate this data for you, anything you write will have to work with the same flawed data that we do. The best decisions are often those designed to keep as many options as possible open until later; deciding when to make the decision rather than making it early with incomplete information. Chris Matts and Olav Maassen have done quite a lot of work on the principles of Real Options which may be helpful for understanding why and how this works.