Hi crab..
if the AP is flying the aircraft you would normally monitor it and let it do its job, but if it does something you don't like or rapid alteration of the flight path is required, the pilot can always move the controls to achieve this without having to disengage the upper modes. Until we get to fly by wire with no physical connection between controls and flying surfaces then this seems the way to go on helos.
NH-90 comes to mind...
I can not comment on machines that are not FBW, but in the FBW machine, I have tended to see more 'instability' when pilots try to 'push against' any upper modes that are engaged, usually quickly followed a flurry of 'uncoupling' of said modes, a mini UA or two before they get back on track.
'Push against' / 'fly-through' technique in ATT made has mixed results in my observation. NHI was nice enough to give us a 'TAC' mode just so we would not have to fly-through. (auto trim follow up) I think its quite neat actually..
As I say, nil experience with other types but in the "1" and "0" machine, I would say does not "like" it. I guess I would describe my own approach as either:
(1) Give "George" long term commands and let him fly,
(2) Manipulate "George" via short term commands and let him fly, or
(3) Boot "George" off (completely de-couple upper modes) and fly ourselves.
But try not to fight George...
Maybe our jobs are evolving toward knowing more about exactly what the upper modes will give us (before we engage it) so we don't get caught with the old "um...whats it doing now?" then having to intervene. In the case of rapid flight path alteration, I have found (on this type) banging off the automatics back to ATT mode tends to work best.
I bring this up because eventually FBW surely will be the standard, bringing us closer to our fixed wing cousins. The future is nearly here!!