archiGo is an experimental architectural proposal, which takes the AlphaGo program (Google’s AI program which plays the board game Go) as inspiration of an intelligent model for decision making. The proposal consists of partial spatial configurations planned as “Tiles”, which continuously assemble to generate larger spatial configurations, with two main objectives being: Fulfil multiple performance goals within both Part to part (Local) relations, and part to whole (Global) relations. The ‘local’ conditions are met through a constraint-solving algorithm, while an AI agent is trained by continuously analysing and evaluating larger assemblages and learning to make improved local tile placement decisions resulting in configurations that better negotiate multiple global goals.