In the domain of Ambient Intelligence, research goals are many times driven by scenarios that help envisage a world enriched by ambient, pervasive, intelligent services. So far, scenarios have most times presented the perception that one person has upon the system, with few details on how the system should work in the background in order to deal with realistic requirements. In this paper, starting from scenarios presented in previous research, we identify features and requirements for AmI systems and propose two new scenarios, focusing on the way information is exchanged beyond the perspective of, and transparent to, the user of the system. We also agentify one of the scenarios, giving insight on how an AmI system may be built, using software agents, in order to fulfill the requirements.