Stochastic resources such as renewables and load uncertainty pose significant challenges for automatic generation control (AGC) of microgrid and interconnected systems. It is challenging to evaluate the performance of AGC systems under non-Gaussian uncertainty due to the high computational burden. This paper presents a novel Ito-theory-based model for the efficient evaluation of AGC performances without time-consuming scenario-based simulations, which is common in conventional approaches. We first present an Ito process model for non-Gaussian stochastic resources with arbitrary probability distributions and then propose a unified stochastic assessment function (SAF) model for AGC constraints and criteria. By expressing SAFs as partial differential equations (PDEs) and leveraging the series expansion of the PDEs, we prove that the SAFs can be expanded into a convergent series of deterministic assessment functions, whose computation is much more efficient than that of the original stochastic problems. Numerical results show the advantage of the proposed approach in AGC assessments under non-Gaussian uncertainty.