Feedback loop

Note:
This aspect of the review is only valid when you are still thinking about human developers.
If you already replaced human developers with AI agents, or if you are planning to do this in the near future, it makes no difference.

Feedbak loop

When a human developer receives the AI review, he can choose to challenge it or accept.
When challenging the review, the developer can point out that the agent's finding was incorrect or that company guidelines are different or something else...
Also, together with AI review, we are running reviews by human reviewers. Their comments could be different than AI.

To accommodate the above, we need another pipeline which will review the reactions from developers or review the human reviewer's input.
Based on these findings, this new pipeline suggests amendments to community/custom/Microsoft rules of the review.
It is also posting short acknowledgements of interactions - like, thank you, noted, or just thumbs up (it is an engaging loop designed to keep the communication with developers open).

In general, it is about a self-learning and self-growing mechanism to improve the quality of review.

Start by reviewing the issue's proposed feedback loop for developer reactions, human reviewer input, rule amendments, and acknowledgements. Define the pipeline boundaries and measurable completion criteria with the maintainers; the issue does not name files, tests, or an implementation entry point.