The Adaptive Keyword Spotting Interaction (AKSI) paradigm presented in this paper is an algorithm for speech command selection in multimodal applications. An architecture implementing AKSI algorithm within IP Multimedia Subsystem (IMS) services design framework is proposed. This architecture uses a SIP/Speech proxy (called Multimodal Enabler) to extract keywords from client applications downloaded to the Mobile Station, based on a set of interaction rules. Keywords are emphasized in the presentation, thus providing an indication on the words to use in speech commands. Synchronization of client application and Speech Interaction Manager is achieved by invoking a call-back method which triggers presentation?s update according to current application context. For unrecognized speech commands, the Multimodal Enabler uses a TTS module to synthesize speech dialogues/prompts which either ask the user to select a command from a given set, and/or to repeat his command.