For inclusive processes of QCD, one naively expects that the quark-hadron duality is at work and the perturbative QCD and OPE can be used to compute the corresponding decay rate or cross sections. But, is the remaining non-perturbative effect totally negligible? How much error do we expect due to the duality violation? Such questions are hard to answer without actual computation using lattice QCD. In this talk, I propose a method to compute the inclusive processes on the lattice. It is non-trivial because one has to include all possible (even unidentified) final states. We apply a technique to compute smeared spectral functions, with which one can bypass the problem of extracting individual final states.