The advent of artificial intelligence in education encourages us to think about what functions we want to automate and what the algorithms will actually learn. If the guiding function is limited to an outcome that is measured within the education system—for example, successful completion of a program as a measure of the success of a tutor—then we may end up with assistants who will prepare for the exam rather than the actual activity. The functions of a mentor, who helps a person to complete a project or solve a professional problem, are more reasonable for automation. But new roles are already emerging, such as an educational broker. Its function is to ensure the most effective investment in human development or arrange a successful trading of your skills in the labor market. Looking at the experience of different organizations we are going to decide which of the roles is the most suitable and ready for automation.