Ambient Intelligence (AmI) refers to an environment that is sensitive, responsive, interconnected, contextualized, transparent, intelligent, and acting on behalf of humans. This environment is coupled with ubiquity of computing devices that enables it to transparently sense context changes, to react accordingly, and even to take the initiative towards fulfilling human needs. Security, privacy, and trust challenges are amplified with AmI computing model and need to be carefully engineered. From software engineering perspective, the shift towards AmI can be seen abstractly similar to the shift from object paradigm towards agent one. Objects provide functionality to be exploited, while agents possess functionality and know how and when to use and o®er it autonomously. Agent paradigm is suitable for implementing AmI considering AmI as an open complex system. Moreover, we argue that agent paradigm is equally useful for engineering all aspects of such systems from the early phases of software development life cycle.