Yeah the biggest practical use of the current generation of LLMs is online astroturphing and viral marketing. AIs aren't reliable enough (yet) for things that require correctness, but they're damn good at saying a thing you want said over and over in countless different ways.

Customer service (aka customer support chat bots) will also try to make use of it, no doubt, but I'm skeptical about how much they'll actually trust it for that. All it'll take is one person figuring out how to make the Chipotle chat bot spout nazi propoganda or some shit before they go back to "dumb" chat bots lol