These past few days I’ve been working on extracting more valuable insights from Jira using AI, with the goal of building useful applications around that data.
I started with Amazon Q, and honestly, the experience was impressive. In a short time, I had an app with a clean interface and a strong PDF report. If you already have access to it, I’d highly recommend trying it. It felt smooth, fast, and surprisingly effective 🚀
Then I was asked to switch to GitHub Copilot, since it is the company-approved assistant.
For clarity, I’m not writing business logic here. My focus is only on retrieving and displaying Jira data.
I began with Copilot in the same way I had started with Amazon Q: I gave it a clear prompt, explained what I needed, and asked it to use a similar app I had already built as a reference.
The first result didn’t work well at all. There were major errors in the code. I kept asking Copilot to fix them one by one, and the code evolved a lot over time. After about 30 minutes, I finally saw the first UI — but it was still far below what I had achieved with Q.
When I tested the app, more issues appeared:
- incorrect functions,
- crashes,
- broken HTML,
- wrong AJAX submission,
- messy JavaScript.
Time for lunch 😅
After lunch, I asked Amazon Q to fix the application and make it behave like the working version in my repo. To my surprise, with a single prompt, I got a clean and working version that matched my expectations.
Smooth. Fast. Stable. 🤖✨
A few reflections
Your mileage may vary, and context matters a lot. Both tools can be helpful. The real difference often comes down to:
- how well the model understands your codebase,
- the quality of feedback you provide,
- whether you have a strong reference implementation,
- and whether you’re willing to switch tools when one gets stuck.
Sometimes switching tools is faster than fighting with a broken conversation.
Most importantly, keep it playful, keep learning, and ship small wins often 😊
Out of curiosity, I also shared Q’s solution with Copilot to see how it would respond. Its answer was surprisingly self-aware and actually quite thoughtful.
One thing I noticed: the major difference may be related to the model version behind the tool. At the time, Amazon Q appeared to be using Sonnet 4, while Copilot was on Sonnet 3.5. Now that Copilot has also moved to version 4, I’m planning to continue the comparison.