Important challenges remain in the development of optimized control strategies for intelligent wells, particularly with respect to incorporating the impact of reservoir uncertainty. Most optimization methods are model-based and are effective only if the model or ensemble of models used in the optimization captures all possible reservoir behaviors at the individual-well and -completion level. This is rarely the case. Moreover, reservoir models are rarely predictive at the spatial and temporal scales required to identify control actions. We evaluate the benefit of the use of closed-loop control strategies, on the basis of direct feedback between reservoir monitoring and inflow-valve settings, within a geologically heterogeneous, thin oil-rim reservoir. This approach does not omit model predictions completely; rather, model predictions are used to optimize a number of adjustable parameters within a general direct feedback relationship between measured data and inflow-control settings. A high-resolution sector model is used to capture reservoir heterogeneity, which incorporates a locally refined horizontal grid in the oil zone, to accurately represent the horizontal-well geometry and fluid contacts, and capture water and gas flow. Two inflow-control strategies are tested. The first is an open-loop approach, using fixed inflow-control devices to balance the pressure drawdown along the well, sized before installation. The second is a closed-loop, feedback-control strategy, using variable inflow-control valves that can be controlled from the surface in response to multiphase-flow data obtained downhole. The closed-loop strategy is optimized with a base-case model, and then tested against unexpected reservoir behavior by adjusting a number of uncertain parameters in the model but not reoptimizing. We find that closed-loop feedback control yields positive gains in net-present value (NPV) for the majority of reservoir behaviors investigated, and higher gains than the open-loop strategy. Closed-loop control also can yield positive gains in NPV even when the reservoir does not behave as expected, and in tested scenarios returned a near optimal NPV. However, inflow control can be risky, because unpredicted reservoir behavior also leads to negative returns. Moreover, assessing the benefits of inflow control over an arbitrarily fixed well life can be misleading, because observed gains depend on when the calculation is made.