Evaluation and adaptation are two open research lines in the field of mobile computing. Due to highly dynamism of the nowadays environments, the software agents running on these complex environments have to face problems of adaptation to user needs. That means that the initial states that prompt the software's decision making process in the first place may dynamically change while the decision making process is still going on just because the user's opinion. So to cope with these problems, such systems should be able to acquire user's opinion and also self-adapt according to it. That is why there is a need of special kind of system that will combine ubiquity, context-awareness, intelligence, natural interaction and adaptation in an AmI environment. Research in context-aware systems has been moving towards reusable and adaptable architectures for managing more advanced human-computer interfaces. In this paper, we assume that the adaptation decisions are taken with the goal of maximizing the user benefit of the applications or services. We estimate their QoS by using the U2E system developed in previuos work and finally propose a methodology for service adaptation.