The paper demonstrates how large-scale stochastic linear programs with recourse can be efficiently solved by using a blending of classical Benders decomposition with a relatively new technique called importance sampling. Numerical results of large-scale problems in the area of expansion planning of power systems and financial planning are presented.