The question is, why bother. Just give the user a chatGPT box where they can ask the AI to do the thing, instead of having to navigate a bunch of web UI.

Conversational interfaces aren’t a great way to rapidly consume information, and they probably aren’t more efficient than a task-focused interface for active interaction in most cases, either. With sufficiently-capable backing model, they seem to be awesome for generality, but that isn’t the be-all, end-all of UI concerns.