The sensemaking process is a complex task in academic research practice. Scholars annotate on their research resources to reflect impressions, insights and facts, as well as reduce the cognitive load in their working memory as a prerequisite for carrying out such sensemaking processes. Given the data-intensive nature of research and the increased collaboration of often interdisciplinary working teams, a vast amount of annotations with a great extent of knowledge are produced daily. However, the annotation tools extant do not usually support sensemaking of those annotations. Using existing research as a foundation, we define a conceptual model for collaborative sensemaking that is based on human-machine collaboration. We introduce concepts to integrate machine intelligence into the collaborative sensemaking process by allowing users to interact with machine-recommended information in favor of exploring knowledge from collective intelligence through interactive visualizations of annotations.