In a behavioural and organisational context, complex problems that reflect the multidimensional attributes of human activity inevitably arise and have to be addressed. In an attempt to model human decision-making behaviour, the vast number of potential parameters raises the question of how this complexity can be harnessed. This paper proposes a data-driven approach, with which dependencies or associations are extracted from the data itself. The complex and dynamic nature of modern business processes makes this approach more suitable, as the design of competent rule-based models or expert systems would be cumbersome, expensive or even infeasible. A hybrid behaviour-modelling method, based on both statistical component analysis and sensitivity analysis, is proposed to directly model decision behaviour. The derived model is the outcome of an optimisation process, where model-data matching is maximised in terms of known, pre-defined criteria. The implementation of the proposed intelligent system as an integral part of real-life business/accounting activity is discussed, and its capability to provide intelligent support to the decision making and internal control processes in management and accounting is demonstrated using realistic data from a business procurement application.