There has been a lot of coverage in recent media about AI being leveraged to expedite vulnerability discovery. On the flip side we don’t see as much about how AI can help us ship better code. At DMS Group, we are often touching code as part of our role, either net new code or more often additions and edits to existing code. Where policy permits, we now conduct AI code reviews for all code we work on as part of our engagements.

If you are responsible for code, this is something we’d commend. Overtime we’ve evolved our own prompts and techniques. A code review using AI models in tool like VS Code can cost as little as 5c and quickly point the team to areas of importance. There are other posts out there that share prompts and approach software developers are using elsewhere but as a start we’d highlight the GitHubCopilot code review approach detailed here.

Of course getting a review is only the start. You still need to review and action the feedback. While its increasingly possible to take an “AI take the wheel” approach we’d strongly recommend against that. If you do not feel comfortable addressing yourself (optionally with AI in support) then loop someone else in. Finally, yes, DMS Group can offer that service (predominantly .NET code). If we can help please reach out.

Happy coding.