Maybe there is no human.
If Elsevier is not a fully robotic scientific publication, somebody ought to create one! Imagine: author uploads the text, Bayesian filter classifies it into an area of knowledge (or the author him/herself does it). The robot then contacts randomly selected candidate reviewers, giving them a unique URL to read the article and to submit the review. Once approved by the majority of reviewers, the article is released for public access.