AI-powered code review assistant

Microsoft's internal AI-powered code review assistant, built in close collaboration with its Developer Division's Data & AI team. It joins each pull request as a reviewer, summarizes the change, and comments on specific lines with suggested edits the author applies. Microsoft says it shaped GitHub's Copilot code review.

Company: Microsoft
Approach type: Agent
Work: Code review
Human involvement: Drafts reviewed
Invocation: Event-driven
Deployment stage: Scaled
Evidence strength: Detailed primary
Entry reviewed: 2026-09-16

Copy as MD Ask ChatGPT

How it works

Pull request creation → AI review → author-applied suggestion.

Automatic first review

Creating a pull request can automatically start the assistant as a reviewer^{[Source 1]}

Review and summarize

The assistant reviews the diff, writes a pull-request summary, and posts categorized findings on specific lines^{[Source 1]}

Suggest changes

For a bug or suboptimal pattern, the assistant proposes a code snippet or alternative implementation^{[Source 1]}

Author decision

The author reviews, edits, and explicitly applies a suggestion; the assistant does not commit it directly^{[Source 1]}

Pull-request Q&A

Reviewers can ask the assistant questions in the pull-request discussion and receive answers about the code^{[Source 1]}

Where people stay involved

People review, approve, or guide the agent’s work here.

Supervision level
Work-product review
A person reviews the draft or implementation.
Level 3 of 5
- pull request → AI review comments^{[Source 1]}

How supervision levels work

Implementation details

Model: Not reported
Harness: Not reported
Sandbox: Not reported
Tool access: Not reported
Knowledge: Repository-specific guidelines and custom review prompts can guide specialized reviews^{[Source 1]}
Context management: Not reported
Credentials: Not reported
Interfaces: Not reportedThe source names the existing PR workflow but does not establish a specific hosting interface.

Validation and failure handling

Reported: Author decision

Reported observations

These results come from the sources. Notes below each result explain what it covers and any reported limits.

Headline claim

Supports more than 90% of Microsoft PRs, impacting over 600,000 pull requests per month^{[Source 1]}

Reported by: Microsoft
Scope: PRs supported by the internal AI review assistant across Microsoft
Denominator: Company pull requests for coverage share

Observed in July 2025

About this result

Category: Adoption output
Evidence type: Reported measurement
What it covers: Pull requests supported across Microsoft per month

Key observation

More than 90% of pull requests across the company^{[Source 1]}

Reported by: Microsoft
Scope: PRs supported by the internal AI review assistant across Microsoft
Denominator: Company pull requests for coverage share

About this result

Category: Adoption output
Evidence type: Reported measurement
What it covers: Share of Microsoft pull requests supported

Key observation

More than 600,000 pull requests impacted per month^{[Source 1]}

Reported by: Microsoft
Scope: PRs supported by the internal AI review assistant across Microsoft
Denominator: Company pull requests for coverage share

About this result

Category: Adoption output
Evidence type: Reported measurement
What it covers: Microsoft pull requests impacted per month

Key observation

About 5,000 repositories in early onboarding^{[Source 1]}

Reported by: Microsoft
Scope: Repositories onboarded in early AI code-review experiments
Method: Early experiments and data science studies across onboarded repositories

The source does not report the denominator of this figure.

About this result

Category: Implementation scale
Evidence type: Estimate
What it covers: Repositories in the early onboarding cohort

Key observation

10–20% median pull-request completion-time improvement in an early 5,000-repository cohort^{[Source 1]}

Reported by: Microsoft
Scope: Median pull-request completion time in an early cohort of 5,000 onboarded repositories
Denominator: 5,000 repositories in early experiments and data science studies
Method: Early experiments and data science studies; comparison design is not specified

Observed in July 2025

About this result

Category: Effectiveness
Evidence type: Reported measurement
What it covers: Median pull-request completion time in the early 5,000-repository cohort

Lessons

Not reported: The reviewed article reports practices and outcomes but this record contains no separate lessons claim.

Back to agents

Improve this record

Give your coding agent the context and instructions to prepare a source-backed update.

Suggest on GitHub

Paste the prompt into your coding agent, then describe the correction or add a public source.

Preview prompt

Help me update the Internal Agents Map record for AI-powered code review assistant at Microsoft.

Repository: https://github.com/steel-experiments/internal-agents-map
Record: data/agents/microsoft-prassistant.yaml
Published record: https://internal-agents.com/agents/microsoft-prassistant
Readable context: https://internal-agents.com/agents/microsoft-prassistant.md

Use the correction or public source I provide in this conversation. If I have not supplied either, ask what I want changed before editing.

Use an existing checkout of this repository, or clone it if needed. Read the current record from the repository. Follow AGENTS.md, CONTRIBUTING.md, and the relevant instructions in data/schema.md.

Verify that the evidence supports the requested change. Preserve uncertainty and distinguish reported facts from catalog judgments. Follow the repository process for capturing new sources. If evidence is insufficient, explain the gap instead of inventing details.

Make a focused change on a branch. Regenerate derived files and run the required checks. Do not edit generated content by hand.

Show me the change, supporting evidence, and check results, including any checks you could not run. Ask whether to submit a pull request unless I already requested one. Do not merge the pull request. If you cannot edit the repository, prepare a source-backed issue draft.

Sources and research details

Citations link to the original publisher. Each source also keeps a preserved copy in the repository, so a changed or removed page stays checkable.

Enhancing code quality at scale with AI-powered code reviewshttps://devblogs.microsoft.com/engineering-at-microsoft/enhancing-code-quality-at-scale-with-ai-powered-code-reviews/Engineering blog · First party · Last source verification: 2026-08-31
Preserved copy

Question coverage and scope

purpose: Reported
workflow: Reported
human involvement: Reported
implementation: Reported
validation: Reported
observations: Reported
lessons: Not reported: The reviewed article reports practices and outcomes but this record contains no separate lessons claim.

Research details for every claim on this page

Summary
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviews
Headline claim
Statement type
Metric
Provenance
Reported
Confidence
Medium
Confidence reason
Microsoft reported the figures in its own engineering blog without independent verification.
Reported by
Microsoft
Scope
PRs supported by the internal AI review assistant across Microsoft
Denominator
Company pull requests for coverage share
Method
Not reported
Observation date
2025-07
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, lines 10
Knowledge
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, line 41
Automatic first review
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, lines 16, 29
Review and summarize
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, lines 18, 23
Suggest changes
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, line 19
Author decision
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, line 19
Pull-request Q&A
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, line 27
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Medium
Confidence reason
Microsoft reported the figures in its own engineering blog.
Reported by
Microsoft
Scope
PRs supported by the internal AI review assistant across Microsoft
Denominator
Company pull requests for coverage share
Method
Not reported
Observation date
Not reported
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, lines 10
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Medium
Confidence reason
Microsoft reported the figures in its own engineering blog.
Reported by
Microsoft
Scope
PRs supported by the internal AI review assistant across Microsoft
Denominator
Company pull requests for coverage share
Method
Not reported
Observation date
Not reported
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, lines 10
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Medium
Confidence reason
Microsoft reported the figures in its own engineering blog.
Reported by
Microsoft
Scope
Repositories onboarded in early AI code-review experiments
Denominator
Not reported
Method
Early experiments and data science studies across onboarded repositories
Observation date
Not reported
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, lines 35
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Medium
Confidence reason
Microsoft reports an early observational result without an independent study or causal design.
Reported by
Microsoft
Scope
Median pull-request completion time in an early cohort of 5,000 onboarded repositories
Denominator
5,000 repositories in early experiments and data science studies
Method
Early experiments and data science studies; comparison design is not specified
Observation date
2025-07
- SupportsEnhancing code quality at scale with AI-powered code reviewsPreserved content.md, line 35
Operating model assessment
Statement type
Inference
Provenance
Catalog judgment
Confidence
Medium
Confidence reason
The engineering blog describes engineers acting on PRAssistant review comments, which locates human attention at work-product review.
Observation date
2025-07
- SupportsEnhancing code quality at scale with AI-powered code reviews