I have created an enemy platform movement AI engine which works really well for one enemy. the problem comes when there is more than one of them on screen. The code is in the main frame in a deactivated group and also as a behaviour for the enemy objects themselves.
You can't use detectors like that either - if you're going to have multiple enemies, you need to use embedded detectors (IMO, you should pretty much *always* be using embedded detectors anyway).