GitHub's agentic engineering guide: commits up 400% YoY, but PR review waits up 441%

GitHub Blog AI/ML · rss · 2026-11-07

GitHub published a developer's guide to agentic engineering, anchored by data showing AI has shifted bottlenecks downstream: over 1.2 billion commits in a single October week (up 400%+ YoY, exceeding all of 2025), while a Faros AI study found PRs spent 441% longer in review for teams moving from low to high AI adoption.

The guide walks through a Plan–Build–Verify–Observe workflow:

Plan

Use Copilot in Slack/Teams to consolidate scattered requirements, generate GitHub Issues with acceptance criteria.
Agentic Workflows auto-label issues and estimate complexity; route low-complexity tasks to Copilot; use Plan Mode for complex design decisions.

Build

Session continuity across devices, cloud sandboxes that keep running after you close your laptop, isolated repo copies per session.
Repository instructions and shared memory persist team conventions (Copilot Dreaming keeps memories current); Copilot Fusion picks models and reasoning depth per task.

Verify

Stacked PRs keep changes reviewable; Code Quality checks standards, AI Scan finds vulnerabilities and unstructured secrets, license compliance gates dependencies.
Agentic Autofix proposes security fixes; Agent Merge resolves failing checks, review comments and merge conflicts.

Original post →

GitHub's agentic engineering guide: commits up 400% YoY, but PR review waits up 441%

More from coding & agent