Your data might be wrong. Much of organizational data is entered by human hands, and humans often make mistakes like typos and spelling variations. Data errors often propagate downstream which can lead to operational errors costing a business time and money. Machines can be trained to recognize where there might be errors in data entry, and can help to correct the data. Machines can also be used to prevent incorrect data from being entered in the first place, by, for example, suggesting likely values and prompting humans to make a simpler, less free form decision. Off the shelf solutions can be cumbersome to configure, and often don't take advantage of the structures unique to your data. In addition, privacy, security and other constraints may prevent you from allowing your data into commingled cloud environments.