Automating report generation text-to-text with a Gen AI component

Hello everyone! I am trying to automate a report generation workflow with no luck. The goal is to get from a 50-60 page word doc to a report that (a) summarizes and curates some of the content using AI, and (b) ensures that all sections of the source doc are covered and correct formatting and doc generation is applied using Python. I’ve set up Python, Langchain, and the openai API on my machine. I am using chatGPT 4o and o3-min-high for code development.

The current workflow is as follows:
Step 1 (Python): Extract sections and generate a structured JSON template.
Step 2 (AI): Generate content for each section.
Step 3 (Python): Format and save the Word document.

Should I keep trying? Or do you see something fundamentally wrong with my approach? TIA