The tremendous increase in customer bandwidth demands has led to the necessity of better utilization of available bandwidth resources. We present an artificial intelligence based pricing mechanism for control of user demand in the obtainable forwarding (OF) service in DiffServ networks. By deploying a fuzzy logic controller at the DiffServ boundary node, we are able to adjust dynamically the price per unit bandwidth, depending on customer bandwidth demands and bandwidth availability. In the case of elastic traffic, a small change in price leads to a small change in customer demand. As the decision to buy bandwidth or not lies in the hands of the customer (his grade of utility), it is difficult to estimate the parameters for the customer bandwidth utility function. The suggested fuzzy controller implements rules, easily obtainable at the boundary node, using terms such as "small change", "big increase". We present an evaluation of the utilization of dynamically changing bandwidth demand.