Generating a self-guided augmented reality (AR) session plan from a remotely-guided AR session held between a remote user and a local user, by: Receiving data recorded during the remotely-guided AR session. Segmenting the data into temporal segments that correspond to steps performed by the local user during the remotely-guided AR session. The steps are detected using at least one of: a Natural-Language Understanding (NLU) algorithm applied to a conversation included in the data, to detect utterances indicative of step-to-step transitions; location analysis of annotations included in the data, to detect location differences indicative of step-to-step transitions; and analysis of camera pose data included in the data, to detect pose transitions indicative of step-to-step transitions. Generating the self-guided AR session plan based on the segmented data and a 3D representation of the scene, the AR session plan including step-by-step AR guidance on how to perform the various steps.