Today, AI can read images, but it would be effective, for example, in contexts where it is embedded in chats as a widget in internal areas to be able to take a snapshot of the user's screen when they ask a question to understand the context they are in and improve service. Being able to give more specific indications as to what he should do or to the question he wants to resolve. Example of an AI that does this: https://www.emy.education/