Audio from user conversations is securely relayed to an OpenAI system, where it is transcribed and answered instantly.