Right now I have: the coloring that there are k-1 subgroups of k-1 vertices. If each of the subgroups contains a connected graph that's one color (like black), and the edges between the subgroups is another color like white, then....
I know that we are trying to show there will be no monochromatic subgraph. So you choose k vertices, and then where do you proceed from here?
You seem to be heading in the right directoin: we take $k-1$ disjoint $(k-1)$-cliques (i.e., complete subgraphs, not just any subgraphs).
By definition, this graph has no $k$-cliques.
The complement is the complete $(k-1)$-partite graph $K_{k-1,k-1,\ldots,k-1}$. Now we just need to argue that any $k$-vertex induced subgraph of this graph is not a clique [this follows from the Pigeonhole Principle].