In this paper, we introduce an architecture for a new generation of speech applications. The presented architecture is based on our previous work with multilingual speech applications and extends it by introducing support for synchronized distributed dialogues, which is needed in emerging application areas, such as mobile and ubiquitous computing. The architecture supports coordinated distribution of dialogues, concurrent dialogues, system level adaptation and shared system context. The overall idea is to use interaction agents to distribute dialogues, use an evaluation mechanism to make them dynamically adaptive and synchronize them by using a coordination mechanism with triggers and transactions. We present experiences from several applications written on top of the freely available architecture.