I'd like to elucidate my opinion because I think I might have made my statement a bit to general for something that deserves a bit more nuance.
Large Language Model AI should not be doing the core aspect of what you are trying to do, whatever that may be. Don't use it to replace yourself, use it to assist yourself, just like any other tool. If you aim to write a story, then ensure that you are the one doing the writing of the story, do not let the LLM write the story for you. You can farm dozens of tasks out to it, but you need to be the story's source.
There are times when maybe you do want an LLM to write for you. For example: writing an assembly manual for an object you designed. If your goal, what you want from your project, is to make and sell x-object, but you know you have a hard time explaining things to others, then it can function as a basic writer.
In essence LLM models, having been trained on massive quantities of data that has been sanitized, produce above average work. Like "good enough" work. If it isn't the main portion of what you are trying to do, then it is able to be value added.
As of right now, however, the things they produce are not standout, and I would wager that with current methods they'll never be better than the pinnacle of human achievment. That would require a radical change in training methods. Many apex examples of human thinking and achievement are practically contradictory among examples of achievment. (Take Quantum Mechanics and Relativity, that each highly accurately describe two portions of reality, but fundamentally contradict each other at their base level of how they're formulated.)
I've rambled for a while, so I'll bring it back in to explain why you shouldn't let the LLM do the thing you want to do. It will come off as rude initially, but there is a harsh kernal of truth in there:
If you use an LLM (AI) to write for you because you feel you yourself are inadequate at accomplishing the task, then someone is more skilled than yourself who also utilizes AI, but as a more effective assistant, will forever outclass you, and you yourself will not build the skill to reach their level.
We are in a brief transitory period where the skilled artists aren't utilizing AI yet, and are having certain jobs taken from them. Some will adapt, and soon will outclass those who tried to use a tool as a standin for their skill.
As an example, while many basic level seamstresses lost their way of life during the industrial revolution, modern fabric and clothing is still designed by industry professionals who merely utilize industrial manufacturing as a tool, rather than replacement, of their skill. This allows products impossible before it was introduced.
In conclusion, integrate llm's into your workflow, just like you use a word processor and keyboard. I storyboard with AI assistance and frequently bounce ideas off it. I ask it for potential ideas (and they're usually shit). It does a decent job of being a sounding board and believability/sanity check. That's where it shines, as a slightly above average feedback machine that you can always ask more details from.