Github Copilot: How to Use AI in the Code Review Process

Recently, my engineering team started experimenting with a new addition to our workflow: asking Github Copilot to review our pull requests (PRs). When you assign Copilot as a reviewer, within minutes, it’ll leave little nuggets of advice or catch sneaky bugs in your code before your teammates even get to it. But how good is it, really? And how should we interpret its code feedback?

After a few months of using it consistently, we’ve learned a lot about what Copilot does well, where it stumbles, and how to get the most value out of its speedy reviews. Through real-world experience and several retrospectives, we’ve integrated Copilot into our code review culture in a way that boosts our speed while keeping quality and human judgement front and center.

In this post, I’ll share how our team has been using Copilot Reviews in practice, and how we’ve built a more collaborative (and sometimes even fun) review process around it!

Table of Contents

• How to Request a Copilot Review
• Where Copilot Shines: Catching the Little…