One of the key concepts in CAS is that the behaviour of the system isn't easy to predict. So if you reach in and poke some element of a complex adaptive system, you can't know for sure what the eventual outcome will be (remember the monkeys around the campfire - was the difference between an opposable thumb and a prehensile tail really so significant??)
And yet every action humanity takes - a single person writing a blog, or a company accidentally spilling oil into the ocean - is poking a CAS on some level. So there are billions of people out there interacting with many levels of CAS every second of every day...And yet by the definition of a complex system, the outcome of all these interactions is difficult to predict.
Giving a monkey a thumb didn't change the world overnight, but over millenia, the ripple effect has been huge. How could anyone have predicted this? Between the monkey and an octopus, I probably would have bet on the octopus being the one to take over the world...After all, it's got eight bloody arms to figure out how to make a good cup of coffee!
Our problem is that in complex adaptive systems, changes are constantly occurring - how do you know which one will have a significant impact? Especially when the impact can take a long time to become apparent?
The evolution of the thumb is a perfect example of a lever point: a small change with a big impact.
Figuring out how to identify the lever points in a given system is a huge motivation for the general study of CAS. I would almost define "self-preservation" as identifying the lever points in the systems with which you interact, then doing your best to push the good ones and prevent the bad ones.