One important step during the conceptual design phase of a new aircraft is to evaluate suitable onboard systems architectures. To analyze and optimize each system in isolation is not recommendable, since it disregards the integration aspects at overall aircraft level and would not promote proper balancing of installation, performance and functional requirements. Therefore, at aircraft conceptual design level, onboard systems should be modelled in a way that enables efficient exploration and comparison of different architectural and technological alternatives. This means that it is important to find a suitable level of modelling and detail to capture the main effects, without adding unnecessary details. Naturally, at later stages more detailed design work will be required, which implies keeping in mind requirements on the facilitation of efficient means for knowledge exchange between different stakeholders within a large organization. To support this approach, it is necessary to be able to explore different architectures relatively quickly, which entails also the capability to efficiently create models of system architectures. To this goal, this paper proposes a novel approach where onboard systems architectural principles are documented by means of ontologies, and where reasoning capabilities offered by a Generative Pre-Trained Transformer is used to automatically extract specific architectural solutions and adapt them to specific requirements. The final goal is to be able to import the resulting onboard system architecture model into the aircraft modelling and simulation tool currently used at Saab.