I have a project going on now that's a proof of concept; working with a client to demonstrate how my company's software can solve their business needs.
Obviously, there are some shortcuts that I can take on this project. For one, I don't have to worry as much about testing because the application is not going to go to production. However, in good agile technique, we are testing during each iteration. At the end, we want to demonstrate a system that works, even though we may not go through all the corner cases we might in a thorough testing approach.
We also don't have to worry about documentation. We'll show the end users how to use the tool, but we don't have to worry about future users. We also don't have to worry about support documentation.
We also didn't do as much up-front work. Typically, because it is a process improvement project, we'll look at the as-is process and figure out the ideal to-be process to implement, then start development based on improvements we've uncovered. In this case, we are just re-implementing the as-is process with the only improvements being what comes with our software; things like improved efficiency, better reporting, better tracking of resources and workloads among other things.
So a proof of concept should be just a first step. Assuming we are successful, we will be able to come in and really help our client. So while the demands on this project aren't as big as a full blown implementation, we still need to show business value. On any project, that's really what it comes down to.