Human and machine collaboration can produce better outcomes than either could achieve alone. An analytical process can be deconstructed into tasks that are better suited for humans or machines. The interface between human and machine is important to design intuitively. An example framework was presented for collaborative data exploration, with phases for data preparation and exploration. Requirements and examples were provided for different task types like category aggregation, binning, clustering, filtering and graph analysis. The key is assigning each step in the process to the agent - human or machine - best equipped to handle it.