Companies that manage to find their way through the maze of network technologies without expert help should not rest on their laurels however, because the real hard work and discipline come with proper data governance. Once a network of sensors have been connected to each other and the cloud, companies can start harvesting that data for later use. But hoarding all the data you can find is not the same as saving it for a useful purpose, a misconception that has spawned a whole family of buzzwords. 'Data swamps' are ungoverned repositories of data that, because of the sheer amount of unqualified data in them, are almost impossible to derive any valuable insight from. AI is only as good as the data fed to it (a concept highlighted in part one called the unreasonable effectiveness of data), and low-quality data can not only ruin the computations of one algorithm, but if the results are trusted and used in an automated process those errors can have devastating real-world implications.