Most high-profile data science wins are deployed as fully automated systems. Consider a recommendation engine. Once deployed, there is no human involved, meaning there is no step in the process for a human to manually replace the algorithmic recommendation with an intuitively-derived alternative. Despite the successes of fully-automated systems, corporations and governments are spending considerable resources to build data science systems to support, rather than replace, human decision making.
In this talk, I will discuss fundamental differences between fully automated and expert-support data science, and best practices around planning, staffing, and proving ROI.