I've just started reading "Agile Project Management" by Jim Highsmith and it's looking very promising. Whilst reading chapter one the point slapped me like the Tango Man - it's the way in which agile projects are different to others.

Traditional project are all about deciding what you are going to build and then carrying out your plan whereas with agile projects you have a plan but it's about how, not what, you build. In fact the key point made by Jim is that agile projects are all about experimentation and adaption.

He cites a comparison with the pharmaceutical industry which makes things very clear. A few years ago they used to sit down and design molecules for a specific task whereas now they synthesise thousands of molecules at a time and use automated testing to screen for desired properties. The results are fed into a database which the scientists use to make their next steps. This rapid experimentation enables them to generate new drugs much faster even though they don't know the end result until it presents itself.

Consider this in the context of software development. Rapid cycles of experimentation with unit tests to screen the results and ensure forward motion. These two concepts enable a project to progress in a more efficient way than if a solution were "designed".