Our Scrum team is debating the ethics and efficiency of using Generative AI (ChatGPT, Gemini) for real-time code reviews. Does it truly help in catching edge cases, or does it just produce generic feedback? We want to speed up our sprints without compromising the technical debt.
3 answers
In the context of Agile and Scrum, Generative AI (ChatGPT, Gemini) acts as a powerful "force multiplier" during the sprint. I’ve implemented it to summarize daily standups and generate technical documentation directly from code commits. Regarding code reviews, it is exceptionally good at finding syntax errors and suggesting optimizations for better readability. However, it can sometimes struggle with deep architectural nuances specific to your local environment. My recommendation is to use it as a "first pass" reviewer to catch the low-hanging fruit, allowing your senior devs to focus on the complex logic and design patterns.
Susan, have you noticed if the team's reliance on Generative AI (ChatGPT, Gemini) has led to a decrease in the collective understanding of the codebase over time?
It’s a huge time-saver for writing user stories and acceptance criteria, which often get rushed during the refinement phase of the sprint.
Laura is right; I’ve started using prompts to generate the "Given/When/Then" scenarios for our stories, and it has made our QA process much more rigorous and structured.
Michael, that’s a valid concern we call "knowledge erosion." To prevent this, we make it a rule that the person using the AI must be able to explain the logic it generated during the peer review. We treat the AI like a junior developer whose work needs a senior’s signature. This keeps the team engaged and ensures everyone actually understands the "why" behind the code changes.