"models degrade in accuracy as soon as they are put in production" - every data scientist relates to this sentence. But why is this such a global truth? and more important - how do you deal with that? In this session, we would like to share some common practices re the data science and engineering activities required in order to maintain the health of the AI/ML models in production, and what can be done in order to identify potential risks as they happen - avoiding the potential damage and mess created while waiting for the feedback loop to come in.