Even when we run the IGA on the same TSP with the same number of clusters, we can get different results because we may not choose exactly the same set of cities for each cluster. Currently, the user manually decomposes the problem and uses the visualization provided by the interface to aid this clustering process. Not only are no two users exactly alike, the same user may not provide the same classification. Without completely getting rid of all interaction we plan to use available clustering techniques to provide an initial clustering that the user can then modify. The automatic clustering will get the same clusters on the same problem each time we run it, the user can then visually adjust these clusters.