Contact center conversations often contain segments with hold music, automatic-recorded-messages or pure silences, where neither the customer nor the agent is speaking. We refer to these segments as Conversational Silences [1]. These silences when continued beyond an acceptable level can negatively impact im portant contact center KPIs, like average handling time, agent efficiency, etc. and may lead to poor customer experience. As a result, it becomes imperative for contact centers to identify si lences in conversations and define mechanisms to better handle them. In this paper, we propose a cascaded system consisting of an ASR engine, a silence detector block, a text classification layer and a heuristic engine to surface instances in calls where agents are missing the protocols to handle silences. This system is used to trigger alerts to agents in real time thus enabling them to course correct while being on call with the customer. More over, these instances can also be surfaced to their supervisors so as to identify agents who are frequently missing these protocols and thereby design dedicated coaching sessions.