The design of cellular systems is a complex, multi-criteria and multi-step process which can have significant implications for the entire organization. Most research in this area focuses on the formation of pan families and associated machine groups, one step in the cell design process. Numerous quantitative techniques have been developed to address this part-family/machine group formation problem. Existing approaches include math programming, algorithms for matrix diagonalization, the application of network modelling and the use of similarity coefficients. These mathematically-oriented techniques can handle a relatively limited set of quantitative objectives and. in addition, require many simplifying assumptions. For this reason, the solutions generated by these techniques are of limited usefulness in actual cell design.This paper proposes an expert system approach to cell system design. The starting point for the expert system is the initial solution generated by traditional mathematical techniques. Based on a flexible set of user-driven quantitative and qualitative factors, the expert system evaluates these preliminary solutions for feasibility and quality. If the solutions are not satisfactory (infeasible or of low quality), the system suggests modifications.