This paper presents an architecture for smart buildings based on common-sense reasoning using the IndoorGML standard. Themain objective is the construction of a knowledge model that allows inferring information that is not explicitly defined in thatmodel. The main challenges addressed are: 1) the modeling of the concepts presented by the standard in the knowledge base used,known as Scone; 2) the automation of the introduction of knowledge in Scone with the proposed model; and 3) the implementationof a service-oriented architecture that makes possible to introduce knowledge in a transparent way. The semantic knowledge wepresent here, which shows more advanced and flexible capabilities, will make it possible in the future to carry out reasonings thatnot only depend on the structure of a building, but on many other aspects of the IoT such as sensors or actuators that are difficult tocontemplate if a common sense approach is not followed.