Disclaimer: in this article, I will talk about Gen AI writing documents, not code.
In my team, we use a JIRA template, which is good because it helps ensure we include enough detail and makes it look nice. But let’s be honest: it takes time to write tickets, so we started to use generative AI to write them. What we started to notice, though, is a drop in attention during refinement sessions. All the tickets started to look (and sound) the same; they were long and hard to read. Therefore, refinement became less useful; people would spend less time discussing the tickets and more time trying to understand them once they were moved into “development”.
In a retrospective, a few team members (including me) highlighted this problem. We decided that we were still allowed to use Gen AI to simplify the writer’s life, but only if it also simplifies the whole team’s life.
The result: people still use Gen AI to write the ticket, but they re-read it and make sure it’s short, sharp and easy to read so we can refine it properly.
Similarly, we noticed that documents such as Request For Comments, Product Requirements Documents and other investigations were becoming long and difficult to read. I mean, 25 minutes to simply read an RFC, who will take the time to review it properly? I also started finding all these documents boring as hell. You scan the document, and in 5 seconds, you can say it was written by an AI. All these docs looked the same to me.
Don’t get me wrong, it’s totally fine to use Gen AI to write documents. It would probably be stupid not to. However, it is important to double-check the content, understand what we write, and make it interesting to read.
The risk of writing these long-form, soulless documents is that people will stop reviewing them and giving feedback. And that defeats the whole purpose of such documents!
As for the JIRA tickets, we agreed to have an extra read-through of the content to make sure it’s clear, concise and doesn’t sound like AI-generated content.
I’ve been writing blog articles on Medium for a couple of years now. I am mainly sharing content about engineering management and leadership.
As English is not my native language, I quickly felt bad that my articles wouldn’t reach the right audience, that people would judge my content, that I was not good enough… So I started to use Gemini to help me initially with the formatting and then to generate a few sentences and sometimes to generate the full draft for me.
I still consider the content as mine, as I spend time finding the ideas, doing the research (and the fact-checking), and I still spend a considerable amount of my personal time on it. However, the result doesn’t “feel” like me. It’s losing its authenticity; it’s too polished, and it gets lost in the flood of AI-generated content.
So, from now on, I have decided to stop using AI to generate and write my blog articles. I will only use it to check the grammar, the spelling and the typos. My content will be less polished, more natural and more authentic. More like me!
If people don’t like it, that is fine. They can go ahead and experience the tone of AI Gen content that’s waiting for them out there. But at least, I will enjoy writing again!
In fact, it’s the first time in a while that I have really enjoyed writing an article. So stay tuned for more “authentic” content!