The present invention is a method and system for reducing bandwidth usage in a voice over internet protocol (VoIP) network that uses holding sounds such as music or announcements. When a customer call enters the network through a gateway, the call is received by a media server (VMS) and a workflow is started. When the workflow determines that the customer call is to be put on hold, the VMS sends an instruction signal to the gateway, instructing the gateway to provide an on-hold music or announcement to the customer call. When the workflow determines that the customer call is to be taken off hold, the VMS sends a second instruction signal to the gateway and the gateway stops the music or announcements and proceeds to further process the call under the direction of the workflow.