A technique is described herein for controlling equipment in a local environment. The technique can use a conversational interface component to receive messages from any user in a group of users. In a media-related context, a media BOT receives each message and determines whether it contains a command to play a particular media item. If so, the media BOT formulates the message as an input query and sends the input query to a search system, with instruction to perform a domain-specific search. The search system identifies a top-ranked media item based on the input query, selected from among a curated media-specific corpus. By virtue of the use of the scoped search system, the technique can successfully leverage preexisting linguistic processing mechanisms. The technique can also generate group profiles associated with users who interact with the conversational interface component, and then use the group profiles to automatically select media items.