You should read the article carefully. In the Introduction section they give some insight for their study:
The majority of AD cases are sporadic in nature. The small fraction of familial cases are caused primarily by mutations in three genes: amyloid precursor protein (APP), presenilin1 (PS1), and presenilin 2 (PS2).
They clearly specify what they are investigating:
Here we demonstrate that a WS extract reverses behavioral deficits and plaque pathology and reduces the Aβ burden in middle-aged and old APP/PS1 mice through up-regulation of liver LRP, leading to increased clearance of Aβ.
So they are investigating the effects of this herb in mice which are "diagnosed" with familial Alzheimers disease (AD).
Less than 0.1% of all cases of AD are familial (Lancet).
The sad thing is that we have no possibility whatsoever to tell whether a patient have the familial type AD. Let´s make a wild assumption that this herb would actually reverse AD in 1% of patients with familial AD (This 1% we would be remarkable in this context). We need to treat 100,000 patients with AD to reverse it in one patient. That kind of "treatment" would not be feasible at all.
Moreover, there are probably hundreds or thousands herbs and drugs which have shown promising results in animals. Unfortunately, >99% of these molecules and potential drugs fail to work in humans. This is due to facts that our fundamental understanding in disease processes are poor and the complexity of the diseases are so overwhelming that the proposed mechanism of recovery observed in the animal subjects with "artificial" diseases just don´t work similarly in human subjects.