Should we support the PR review process with AI?

[anonymoususer72041]

I would like to open a discussion about whether we should consider using AI tools to support the review process for pull requests in OpenCATS.

The main motivation would be to make PR reviews faster and more consistent, especially for common checks such as code style, possible regressions, missing tests, unclear logic or documentation gaps. An AI-assisted review could help maintainers and contributors get earlier feedback before a human reviewer spends time on the PR.

At the same time, this would clearly not replace human review. AI-generated feedback can be wrong, incomplete, misleading or too confident. There is also a risk that maintainers may spend extra time verifying incorrect suggestions or that contributors may feel discouraged by automated comments that are not actually useful.

Some possible benefits could be:

• Faster first feedback on pull requests
• Earlier detection of simple issues
• More consistent review coverage
• Reduced workload for maintainers on repetitive checks
• Helpful suggestions for contributors who are less familiar with the project

Some possible downsides could be:

• False positives or incorrect recommendations
• Additional noise in PR discussions
• Risk of relying too much on automated feedback
• Extra maintenance effort for configuring and monitoring the AI review process

If we decide to experiment with this, I think it would be important to keep the process conservative. For example, AI comments should be treated as suggestions only, not as required changes. We could also start with a limited trial, make sure comments are clearly labeled as AI-generated, and define when maintainers should ignore or remove unhelpful feedback.

I would be interested in hearing what others think about this.

Would AI-assisted PR reviews be useful for OpenCATS?

Are there specific tools or workflows we should consider?

What rules or limits would we need to avoid creating more noise than value?