The field of virtual reality (VR) has undergone significant development these last years due to the growing maturity of this technology. This has allowed for its diffusion in many domains, notably in the building construction one. However, a problem resulting from such use of VR is that the VR interactions provided to the users are often not sufficiently adapted to their needs. It is indeed difficult to make a synthesis between the building trades’ expertise and the VR expertise. A research question then arises, which is how to allow the experts in a specific field, who do not have VR expertise, to be able to define and obtain the specifications of VR interaction technique, in autonomy. To answer this question, in this research we propose a new task-oriented methodology for the design of VR interactions. This one contains semi-automated systems that allow for the decomposition of the user task and for the determination of proposals of VR interaction techniques - i.e. the obtaining of the two types of specifications that are expected here. This methodology has been tested and evaluated on two case studies from the building sector. The results obtained show that our methodology can be used by experts from the construction industry, in autonomy, and that they have obtained specifications similar to the ones obtained by following a traditional user-centred design methodology.