no its not just asking an llm to handle the process. good models can make acceptable output but the cost is nearly 100x. bad models just hallucinate text. i do expect this gap to close in the coming years though.