In consulting "AI" firms, do you think engineering is more necessary than the actual AI part? To me, it seems that in those situations, knowing when and where to apply certain models could actually solve a lot of business problems for a client.