Using ChatGPT to Draft Project Documentation

I write a lot of documentation. Project charters, sprint summaries, stakeholder updates, risk registers. It is part of the job. Three months ago I started using ChatGPT with GPT-4 to draft these documents, and it has meaningfully changed my workflow.

Let me be clear about what I mean by "draft." I do not paste a prompt and ship the output. I use GPT-4 as a first-draft engine. I give it context about the project, the audience, and the format I need. It generates a structure and initial content. Then I rewrite, correct, and add the details only I know.

Where It Works Well

Status reports and sprint summaries are the sweet spot. I paste my notes from the sprint, specify the audience (leadership, client, team), and ask for a structured summary. The output is usually seventy percent usable. I spend ten minutes editing instead of thirty minutes writing from scratch.

Meeting agendas are another win. I describe the meeting purpose and attendees, and GPT-4 produces a reasonable agenda with time allocations. It is not perfect, but it gives me a starting point that is better than a blank page.

Where It Falls Short

Anything requiring deep project context is still manual. Risk assessments, architectural decision records, and stakeholder-specific communications need nuance that the model does not have. It does not know your stakeholders' personalities or your project's political landscape.

I also never use it for anything containing sensitive client data. The information governance question is still unresolved at most organizations, and I am not going to be the person who leaked a client's roadmap into a training dataset.

My Advice

If you are a PM and you have not tried this yet, start with low-stakes documents. Build a prompt library for your recurring formats. Treat the AI as a junior writer who needs heavy editing. The time savings are real, but only if you maintain quality control. The human in the loop is not optional.