Our first application of Agile principles to the research process, which occurred during our study of thought leadership in the tech industry, was a very Agile-esque journey into the unknown. We learned a great deal that applies to future applications of Agile Research Development, so here's my retrospective.
You Can Be Agile Up To The Boundaries Of What You Control
Ultimately, you control only part of the value stream. This maxim holds true wherever you apply Agile, and software developers learn this principle right away. You can be the most disciplined developer imaginable, never checking in code that doesn't work, always building in tests, and still be at the mercy of your own QA team. And even if the larger team, including testers, works in blissful harmony, someone needs to deploy the code on a live system. (Which is why dev ops is receiving a lot of attention from within the Agile community these days. See Jez Humble's recent book for a good example.)
If a stream is an appropriate metaphor for value creation and delivery, then this stream always takes many twists and turns, frequently encountering dams and obstructions. The success of Agile, therefore, depends to a great extent on eliminating these kinks, navigating around them, or learning to accept them as part of the value stream's trajectory.