Order of the Phoenix
GitHub Unveils a Four-Stage Guide to Agentic Engineering with Copilot
GitHub has published a guide to agentic engineering with Copilot, urging developers to let agents aid every stage: Plan, Build, Verify and Observe.
Monday, 9 November 2026 · 1 min read
GitHub has published a guide to what it calls agentic engineering, a way of working in which AI agents, chiefly Copilot, assist developers across four stages: Plan, Build, Verify and Observe. It matters, the company contends, because as AI speeds up implementation, developers feel more pressure in other places, most notably in the weary queues of code review. Behold the figures GitHub sets before its readers. According to GitHub, it saw more than 1.2 billion commits in a single week this October, over 400% more than a year earlier and more than it received in all of 2025. GitHub also cites a Faros AI study which found that pull requests spent 441% longer in review when teams went from low to high AI adoption. For planning, GitHub counsels using Copilot in Slack or Teams to gather scattered context, then asking it to create GitHub Issues with acceptance criteria. GitHub Agentic Workflows can label issues and estimate task complexity, while Plan Mode lets developers compare approaches and approve a plan before Copilot begins. For building, the guide describes cloud sandboxes, session continuity across devices, Copilot Dreaming to keep shared memories current, and Copilot Fusion to choose models and reasoning depth. The verification chapter is perhaps the most arcane. GitHub recommends stacked pull requests, meaning dependent work split into smaller pieces so that each can be reviewed on its own. It further enlists GitHub Code Quality, Copilot code review, AI Scan, license compliance checks and Agent Merge, which, by GitHub's account, automatically addresses failing checks, review comments and merge conflicts. Finally, for observing software once shipped, GitHub says Copilot can draw on observability tools and business data through connectors such as Fabric to help investigate which requests or customer groups are affected by a fault. Copilot may then draft an issue complete with a regression test. These are GitHub's own claims about its own wares, which The Chronicle has not independently tested.
Source: github.blog