At the present time several powerful simulation models exist for the assessment of building environmental performance at the design stage. However when used as design tools these models suffer from several fundamental limitations. Typically they fail to tackle the problematic issues surrounding data preparation in the face of uncertainty. Invariably models are functionally orientated, containing little knowledge of the application domain. This means that they cannot direct a users' line of enquiry, allowing 'Why do you ask' type responses for example. Instead, the user must be expert enough to devise an appropriate performance assessment methodology and to coordinate model operation against this. These are the limitations being addressed in a project to develop an Intelligent Front End (IFE) for building performance simulation. The objective is the design of a machine environment which can act as an expert consultant to assist the user in the problem description phase, recognise the user's appraisal wishes, commission simulations and report back on overall performance. This paper describes the form and content of the IFE which is conceived as an intricate synthesis of user modelling, human-computer interaction techniques, contextual knowledge and the interface to the various performance prediction models at its back-end. It also discusses the knowledge entities contained within, and manipulated by, the IFE.