Supervised learning

Learning from examples someone has already labelled. You supply a few thousand invoices with the total marked, or emails tagged spam and not spam, and the system works out the rule that separates them.

Use it when you know what the right answer looks like and you have enough past cases to show it. That covers most business automation: classifying incoming requests, reading documents, flagging anomalies, predicting a number from history. The catch is the labelling. Somebody has to have done it, and if the labels are wrong the system learns the mistake perfectly.