Your CI/CD pipeline runs linters, tests, and deploys. But it doesn’t understand your code. Adding AI to the pipeline gives you automated code review, intelligent test generation, and deployment risk assessment — without a human in the loop.

AI code review in GitHub Actions

# .github/workflows/ai-review.yml
name: AI Code Review
on: [pull_request]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0

      - name: Get PR diff
        run: git diff origin/main...HEAD > diff.txt

      - name: AI Review
        run: |
          REVIEW=$(curl -s https://openrouter.ai/api/v1/chat/completions \
            -H "Authorization: Bearer ${{ secrets.OPENROUTER_KEY }}" \
            -H "Content-Type: application/json" \
            -d "{
              \"model\": \"openai/gpt-5.4-mini\",
              \"messages\": [{
                \"role\": \"user\",
                \"content\": \"Review this code diff for security issues, bugs, and performance problems. Be specific and actionable. Format as a GitHub comment with markdown.\n\nDiff:\n$(cat diff.txt | head -500)\"
              }]
            }" | jq -r '.choices[0].message.content')
          
          # Post as PR comment
          gh pr comment ${{ github.event.pull_request.number }} --body "$REVIEW"
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Cost: ~$0.01-0.05 per PR review with GPT-5.4 Mini. For free, use Qwen 3.6 Plus via OpenRouter.

AI-generated PR descriptions

- name: Generate PR Description
  run: |
    DIFF=$(git diff origin/main...HEAD | head -1000)
    DESC=$(curl -s https://openrouter.ai/api/v1/chat/completions \
      -H "Authorization: Bearer ${{ secrets.OPENROUTER_KEY }}" \
      -H "Content-Type: application/json" \
      -d "{
        \"model\": \"openai/gpt-5.4-mini\",
        \"messages\": [{
          \"role\": \"user\",
          \"content\": \"Write a concise PR description for this diff. Include: what changed, why, and any breaking changes.\n\n$DIFF\"
        }]
      }" | jq -r '.choices[0].message.content')
    
    gh pr edit ${{ github.event.pull_request.number }} --body "$DESC"

AI deployment risk assessment

Before deploying, let AI assess the risk:

- name: Deployment Risk Check
  run: |
    CHANGES=$(git log --oneline origin/main...HEAD)
    FILES=$(git diff --name-only origin/main...HEAD)
    
    RISK=$(curl -s https://openrouter.ai/api/v1/chat/completions \
      -H "Authorization: Bearer ${{ secrets.OPENROUTER_KEY }}" \
      -H "Content-Type: application/json" \
      -d "{
        \"model\": \"openai/gpt-5.4-mini\",
        \"messages\": [{
          \"role\": \"user\",
          \"content\": \"Assess deployment risk (LOW/MEDIUM/HIGH) for these changes:\n\nCommits: $CHANGES\n\nFiles changed: $FILES\n\nRisk factors: database migrations, auth changes, API breaking changes, config changes. Respond with RISK_LEVEL: LOW/MEDIUM/HIGH and a one-line explanation.\"
        }]
      }" | jq -r '.choices[0].message.content')
    
    echo "$RISK"
    if echo "$RISK" | grep -q "HIGH"; then
      echo "::warning::High-risk deployment detected. Manual approval recommended."
    fi

AI changelog generation

- name: Generate Changelog
  if: github.ref == 'refs/heads/main'
  run: |
    COMMITS=$(git log --oneline $(git describe --tags --abbrev=0)..HEAD)
    
    CHANGELOG=$(curl -s https://openrouter.ai/api/v1/chat/completions \
      -H "Authorization: Bearer ${{ secrets.OPENROUTER_KEY }}" \
      -H "Content-Type: application/json" \
      -d "{
        \"model\": \"openai/gpt-5.4-mini\",
        \"messages\": [{
          \"role\": \"user\",
          \"content\": \"Generate a user-facing changelog from these commits. Group by: Features, Fixes, Improvements. Skip internal/CI changes.\n\n$COMMITS\"
        }]
      }" | jq -r '.choices[0].message.content')
    
    echo "$CHANGELOG" >> CHANGELOG.md

Using local models in CI (self-hosted runners)

If you use self-hosted GitHub Actions runners, you can use Ollama for free AI in CI:

jobs:
  review:
    runs-on: self-hosted  # Your runner with Ollama installed
    steps:
      - name: AI Review (local, free)
        run: |
          git diff origin/main...HEAD | ollama run qwen3:8b "Review this diff for bugs:"

Zero API cost, zero data leaving your network. See our self-hosted n8n guide for the infrastructure setup.

Cost per pipeline run

Task	Model	Cost per run
Code review	GPT-5.4 Mini	~$0.02
PR description	GPT-5.4 Mini	~$0.01
Risk assessment	GPT-5.4 Mini	~$0.005
Changelog	GPT-5.4 Mini	~$0.01
All four		~$0.05
All four (local Ollama)	qwen3:8b	$0

At $0.05 per PR, a team doing 20 PRs/day spends $1/day on AI-powered CI. That’s cheaper than one missed bug in production.

Related: AI App Deployment Checklist · Deploy AI Agents to Production · Claude Code Routines · Self-Host n8n with Local AI · LLM Regression Testing · OpenRouter Complete Guide