Using retime is one option, you would just have to shape your array into a timetable first, which is a bit of a hassle.
I think this should work as well
%some data
A = 1:10000;
A = reshape(A,10,10,100);
t = (linspace(datetime(2000,1,1),datetime(2005,1,1),100))';
%find unique month ids
[U,~,G] = unique([year(t),month(t)],'rows');
%preallocate output variable
out = nan(size(A,1),size(A,2),size(U,1));
%loop over months
for i = 1:size(U,1)
f = A(:,:,G==i);
out(:,:,i) = mean(f,3);
end
