Multi-agent conversational systems are multi-layered AI solutions that implement multiple intelligent agents working toward a common objective. Unlike single-agent chatbots, these systems are composed of different agents who perform various roles, such as answering questions or performing certain tasks.

The essential elements that constitute such systems encompass dialogue management (so that the conversation is threaded together), natural language understanding (NLU) (to make sense of the user’s utterances), and response generation (to provide suitable context for the replies).

https://dzone.com/articles/building-multi-agent-systems-with-autogen-framework