Convert individual prediction standard errors into aggregate error?

Question

I have standard errors for each prediction from a Tweedie GLM, is it possible to leverage this to obtain the standard error for an aggregate sum of predictions for some cate