Although you can add new documents to a learnset at any time, you may find inconsistent results when you add new documents to a learnset and then save the project without learning. This only happens when you use a project that has a history of learning, as the classification engine has a knowledge base from which it extracts information.
Without proper learning, the classification does not include the information from the new documents. To remedy this, change the learnset by manually adding a document to it and then changing the working mode. The documents automatically go through the learning process once the mode changes.
Use this method for small projects and demonstration purposes only. For large projects in which learning is a lengthy task, see About Learning