I find it easier to understand without focusing on electron shells/unpaired electrons.

When a magnetic material is run through an AC coil, it is magnetized in one direction and then flipped 180 degrees (both coincident with the axis of the coil).  As you pull the material out of the coil, reverse magnetic domains are formed.  The reverse domains tend to magnetically couple to their neighbors.  This causes a net reduction in magnetic output of the material.

When a magnetic material is heated above the Curie point and then cooled, numerous magnetic domains spontaneously form in the material.  Each domain magnetically couples to its neighbors.  The net effect is zero magnetic output, although at the domain level, each domain is fully magnetized.

Another effect of the AC coil, as the magnetic domains are flipping back & forth, the part is moving to areas of diminishing field strength so eventually the residual magnetism approaches zero.

A magnet that has been AC demaged isn't demaged on the microscopic level, but on the macro level.  By flipping domains you end up with then both less than fully saturated and with random orientations.  Combined this results in a near zero total field.

= = = = = = = = = = = = = = = = = = = =
Plymouth Tube

Just a question to Ed:  Within a domain, aren't all the magnetic moments aligned?  If so, a domain could be considered a self-contained saturated magnet.  In a demagnetized magnet, all the domains would be randomly oriented (along the preferred crystallographic orientation), and since they magnetically couple to their neighbors, there is no net magnetization.

I've always envisioned a magnetic domain implies a fully saturated region within a magnet.

Not all of the time.  When the magnet is saturated they all are aligned.  After all you are not adding/removing electrons, but prior to saturation there is no field even on a micro level.
I say this because if you take a magnet as heat treated you get a very weak response using a magnetosensitive tint etch.  This is much different if you do it to a demagnetized part.  Even an alloy like Alnico which can be well demaged.

= = = = = = = = = = = = = = = = = = = =
Plymouth Tube