Disclosed are apparatus and techniques for escalating an instant messaging (IM) session between two or more users to an audio session so that a communication device (by one or more users) that will optimize the audio session is selectively and intelligently utilized. In one embodiment, if an IP (Internet Protocol) phone is connected to the computer on which an IM session is being escalated, the IP telephone may be used instead of the computer device's own soft telephone so as to maximize audio clarity and quality. In another embodiment, a user's cell phone may be utilized for a mobile user who is using a laptop that is not coupled to an IP telephone or who is communicating from a hotel room in which a telephone call is typically more expensive than a cell phone call. Numerous embodiments for optimizing such audio sessions are described herein.