Personally, my (uneducated) opinion is that we already have plug-and-play functionality on a program level ie I can add an OpenAI api key to various programs and make them 'smarter'. Since the Linux experience is often pretty piecemeal as is, this would be a solid enough approach for most.
In terms of AI being ingrained within a Desktop Environment, that seems harder for me to imagine... Like how the Office Suite has AI functionality, would the KDE suite of apps allow for cross-program functionality? Would this require a substantial change in system requirements for local processing? Would there be an open-source LLM hosted in the cloud for chat purposes that also mirrors the privacy expectations of the average Linux user?
I understand people's apprehension towards Linux distros seemingly chasing the latest fad, but I think it's also worth hypothesizing the alternative if AI and LLMs are here to stay/differentiate.