In conventional microscopy, complex microscopy methods or detection of rare event needs supervision. The microscopist is in front of the microscope to choose the adequat cell and/or to set up the acquisition. On the other hand, biology needs high-content screening (HCS) to understand the complexity of live, microscopy using HCS are usually only basic unsupervised acquisition. Smart automated microscopy will open new avenue to combine complex acquisition or detection of rare event with HCS in an unsupervised manner. The Roboscope project aims to develop sequence-driven acquisition on an automated fluorescence microscope by integrating real-time image analysis by artificial intelligence to implement feed-back loop. The objective is to allow unsupervised acquisitions of advanced microscopy methods both to track rare events or to acquire high through-put. This collaborative project involves the teams of Jacques Pécréaux and Marc Tramier at the IGDR with the company Inscoper. The technological choice of our project to achieve the real-time constraint is to develop embedded solutions for image acquisition, image analysis, and feed-back loop automation. In this workshop, we will show how an embedded AI algorithm enables real-time execution in an automated microscopy servo sequence. We will then apply it to detect automatically mitotic cells from a mosaic acquisition of chromatin label in order to choose the best cell to follow dynamics of mitotic spindle using a second fast sequence with another objective and another color.