Smart cities will need collections of buildings that are responsive to the variation in renewable energy generation. However, an unprecedented level of renewable energy being added to the power grid compounds the level of uncertainties in making decisions for reliable grid operation. Making autonomous decisions regarding demand management requires consideration of uncertainty in the information available for planning and executing operations. Thus, this paper aims to quantitatively analyze the performance of supervisory controllers for multiple grid-integrative buildings with thermal energy storage depending on the quality of information available. Day-ahead planning and real-time model predictive controllers were developed and compared across 50 validation scenarios when given perfect information, deterministic forecasts, and stochastic forecasts. Despite the relatively large uncertainty in the stochastic forecasts, marked improvements were observed when a stochastic optimization was solved for both the day-ahead and real-time problems. This observation underscores the need for continued development in the area of stochastic control and decision-making for future grid-interactive buildings and improved energy management of smart cities.