This work presents a model of organisational problem solving able to account for the relationships between problem complexity, tasks decentralization and problem solving efficiency. Whenever problem solving requires the coordination of a multiplicity of interdependent elements, the varying degrees of decentralization of cognitive and operational tasks shape the solution which can be generated, tested and selected. Suboptimality and path-dependence are shown to be ubiquitous features of organisational problem solving. At the same time, the model allows a precise exploration of the possible trade-offs between decompostion patterns and search efficiency involved in different organisational architectures.