Quality assurance · Production

Microsoft's AI-powered code review assistant scales to over 90% of PRs and 600K pull requests per month

The problem

PR reviews at Microsoft had significant friction: reviewers spent time on low-value feedback while meaningful concerns like architectural decisions and security implications were overlooked, and at scale—thousands of developers and repositories—PRs waited days or weeks before being merged.

Workflow diagram · grounded in source
1
PR creation triggers AI review
trigger
“Whenever a pull request is created, the AI assistant automatically kicks in as one of the reviewers”
2
AI flags code issues with categories
ai_action
“AI reviews the code changes and leaves comments just like a human reviewer would. It flags a range of issues – from simple things like style inconsistencies and minor bugs, to more subtle concerns like a potential null reference or an in…”
3
AI suggests code improvements
ai_action
“the assistant even suggests specific code improvements. If it identifies a bug or a suboptimal code pattern, it proposes a corrected code snippet or an alternative implementation for the author to implement”
4
Author reviews and applies suggestions
human_review
“The author remains in control—reviewing, editing, and deciding whether to accept the suggestion by explicitly clicking 'apply change' option. All changes are attributed to the commit history, preserving accountability and transparency.”
5
AI generates PR summary
ai_action
“AI also generates a summary of the PR – essentially an AI-written description of what the code change is doing. The AI looks through the diffs and tries to explain the intent of the change and highlights key changes.”
6
Interactive Q&A with AI
ai_action
“Reviewers can also engage the assistant in a conversation within the PR discussion. If something in the code is unclear, a reviewer can ask the AI questions about the code or request clarification.”
Reported outcome

The AI reviewer now supports over 90% of PRs across Microsoft, impacting more than 600K pull requests per month, with repositories onboarded to the tool observing 10–20% median PR completion time improvements.

Reported metrics
PR coverage across companyover 90%
Pull requests impacted per month600K
median PR completion time improvement10 – 20%
Reported stack
AI code reviewerlarge language models
Source
https://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The AI reviewer now supports over 90% of PRs across Microsoft, impacting more than 600K pull requests per month, with repositories onboarded to the tool observing 10–20% median PR completion time improvements.

What tools did this team use?

AI code reviewer, large language models.

What results were reported?

PR coverage across company: over 90%; Pull requests impacted per month: 600K; median PR completion time improvement: 10 – 20% (source-reported, not independently verified).

How is this quality assurance AI workflow structured?

PR creation triggers AI review → AI flags code issues with categories → AI suggests code improvements → Author reviews and applies suggestions → AI generates PR summary → Interactive Q&A with AI.

Related quality assurance cases