AI in Tech Writing: A Case Study

I was hired by a project manager at a client to write end user documentation for a new system the company had implemented. As part of the engagement, I obtained consent to use AI to accelerate the documentation process. We developed an effective workflow:

1. Record a demo: The PM would record a demo of a specific feature of the system’s UI.

2. Create a transcript: I then extracted the audio from the demo video and uploaded it into my Office 365 account. Word 365 includes a built-in transcription feature (it’s reasonably accurate). From the audio file, I generated a full transcript.

3. Generate a draft: I uploaded the transcript into ChatGPT and asked it to create a procedural document. The result was well-structured, readable, and serviceable as a first draft.

4. Human editing: Because software demos rarely narrate every mouse click or keystroke, I watched the demo video while editing the AI-generated draft. I paused, rewound, and added details to ensure the procedure was complete and accurate.

What would normally have taken eight hours per video dropped to about three. Across ten demo videos, this was a substantial efficiency gain.

I realize there are transcription tools for chats (e.g., Otter.ai) and alternatives to the process I’ve described. Not all companies have embraced them and I find the trick these days is to choose tools and methods acceptable to the client. Indeed, I’m finding the real issue for the use of AI in business practice overall is data security.

David Richards