Computation of QCD processes is always hard. The use of lattice QCD is limited to low-lying hadronic (often single-particle) states, so that only exclusive processes can be computed. Perturbation theory, plus operator product expansion, is used for high-energy processes for which the final states are not specified, or summed over all possible states, i.e. inclusive processes. There are annoying cases, where two analyses are mutually contradictory, as seen in the |Vcb| and |Vub| determinations. They suggest that there remain some unknown systematic effects in either or both analyses. In this talk I argue that one can construct a theoretical method that can treat both cases in principle, by extending the lattice QCD approach.