Usually most professions have a direct or at least are stakeholder led QA function that monitors their output, however there are professions that tend to lack this kind of QA, such as architects, doctors and other professions that usually involve the professional to confess specialised knowledge in a wide area, so that testing the output is often really impossible. Usually with traditional professions there is a system of self governance were the professional needs to belong to a society that ensures this.
The problem with IT architects in their many variances from business architects to solution and software architects is that we are lacking this kind of society and we usually are also lacking feedback control as to the complexities involved. Usually the best way to QA an architect is therefore to deploy random intervallic reviews by experienced architects. Here it is important to engage a reviewer that is independent enough and at the same time knows enough of the subject matter, which often turns out to be impossible. Therefore reviews usually focus on formalities that however are often not agreed in many companies.
Therefore for architecture reviews to work the architecture manager need to insist on the architects designing and agreeing on an architectural review process and then also establishing the architecture review resources ideally from an outside perspective.