Here is my method:
1) I use back-button auto focus control all the time. That means my shutter button and my focusing button are separated.
2) I use continuous AF mode (AI servo) all the time, the only exception is in dark conditions when I need the a speedlite flash AF assist light (because that won't enable on any mode other than one-shot)
3) I typically, but not exclusively, use "all the time manual focus lenses (USM, HSM etc...)
With the above I am able to point the camera at what I want and enable the af when I want to by just pressing the button on the back of the camera. This engages AF (and IS if I have that on the lens and enabled) and thus the camera starts focusing and tracking the focus. If I want to treat the AF like one shot I just enable the AF with the back button till it focuses on what I want then release that button. I can then take the shot as normal and the AF won't engage when the shutter button is pressed, thus leaving me free to do what I want without having to always be switching modes.
When using all the time manual focus lenses I can go a step further, by not pressing the backbutton the AF won't engage; thus I can use the focus wheel on the lens and focus how I want and take the shot; again the AF won't kick in when the shutter is pressed. Thus I can shoot in one shot - continuous and manual focus all without changing a single setting.
But there's more, if I'm using a lens with IS/OS (or VR if you're Nikon) then I can half press the shutter button to engage and spin-up the anti-shake. This is great if I've pre-focused a shot or for whatever reason, don't want the AF to engage at that point in time. This means I can "point and shoot" an action shot and my anti-shake is already spun up and already working and ready for the shot when I want it to be.
The only time I really have to change into one-shot AF mode is when using a speedlite flash and wanting to have the AF assist light illuminate the target; because that won't engage with continuous AF modes.
I avoid AI Focus because the camera doesn't know what its looking at nor can it predict when action might happen; I do (or least ways have more chance than the camera) so I don't let it do that bit of thinking for me. I do it and that way I get a reliable, repeatable and known result when I press the back button to engage the AF.