Should AI automatically triage GitHub Issues?

[hemathens]

🏷️ Discussion Type

Question

💬 Feature/Topic Area

Issues

Body

Imagine GitHub automatically assigning labels, priorities, duplicate detection, and suggested assignees.

Would you trust it, or would it create more work than it saves?

Guidelines

• I have read the above statement and can confirm my post is relevant to the GitHub feature areas Issues and/or Projects.

[nigusmamo]

I would trust it as a starting point, but not completely. AI can save a lot of time by suggesting labels, priorities, and possible duplicates, especially in large projects. However, I think maintainers should still review the suggestions because context matters, and incorrect assignments could create extra work.

[EgglezosHub]

I think a hybrid approach is the sweet spot! Having AI do the heavy lifting in the background (like scanning for duplicates and suggesting labels) would save maintainers a massive amount of time, especially on big open-source projects. But it should definitely require a human to click approve before any labels or assignees actually go live. If the AI makes a wrong guess and hides a critical bug or pings the wrong contributor it ends up creating more messy cleanup work than it saves

[farhan2t24]

I think it would save time, especially for repositories with a high volume of issues. Automatic labeling, duplicate detection, and suggested assignees could help maintainers triage issues more quickly.
That said, I wouldn't want it to be fully automatic. AI can misclassify issues or suggest the wrong assignee, so maintainers should be able to review and override its recommendations easily.
My preference would be AI-assisted triage with human confirmation rather than AI making final decisions. That provides most of the productivity benefits while keeping errors under control.

[Reznouw]

I would trust AI triage as an assistant, not as the final decision-maker.

It could be very useful for first-pass work like:

• Suggesting labels.
• Detecting likely duplicates.
• Estimating severity or priority.
• Recommending related issues or PRs.
• Suggesting possible assignees based on ownership/history.

But I would be cautious about making those changes automatically without review. Issue triage often depends on project-specific context, roadmap priorities, maintainer availability, and subtle details that AI may miss.

The best version, in my opinion, would be AI-generated suggestions with easy human approval, editing, or rejection. Maybe repositories could choose different modes:

• Suggest only.
• Auto-label low-risk categories.
• Require approval for priority/assignee changes.
• Fully automatic only if the maintainers explicitly enable it.

So yes, I think AI triage could save a lot of time, especially for large repos, but it should be configurable and transparent. Maintainers should always be able to see why a label, priority, duplicate, or assignee was suggested.