I hadn't used accumarray before, so due to the comment by @Dan I decided to give it a try.
At first I tried a naive version and used histc to count occurrences to get the desired mean values... (Note that accumarray will sort the output the same order as unique, so mean will be calculated correctly)
%// Naive version
ua = unique(A(:,1)); %// use as histc bins (or sorted "group" values)
result = accumarray(A(:,1), A(:,2)) ./ histc(A(:,1), uA);
Here, accumarray operates by summing over all entries in A(:,2) corresponding to identical subscripts in A(:,1).
But then I realised that by passing the optional fun argument to accumarray changing "summing" to "mean", you can do this as a one-liner:
%// one-liner
result = accumarray(A(:,1), A(:,2), [], @mean);