This is a very interesting point. The whole goal behidn "time saving" devices is that the time spent setting them up is amortized over each successive use. So, let's say you have a process that takes 10 seconds. If you can cut it down to 5 seconds, but take 50 seconds to set up that new process, you need to do your process more than 10 times to see any savings. If you do it less than 10, you actually have lost time.

This is a very big part of XP programming, particularly YAGNI. If you only do the simplest thing that could possibly work, then you haven't spent time that may not be amortized. However, if you do something 2-3 times and can reasonably expect that it will be done more times (like different reports, similar classes, etc), then you should spend time setting up a framework because that time will be amortized in the short-term.

However, especially outside programming, figuring out the cost-benefit ratio of time-saving devices is very tricky because we don't know how many more times we will need to run that process. Washing machines are easy, but is a leaf-blower?