Agentic engineering for teams: What changes when every engineer runs agents

A.Team, which supplies engineering teams, advises that an AI agent's pull request needs an approver other than the engineer who asked for it, because review breaks first.

A.Team, a company that supplies engineering teams to other companies, published a guide on what changes when every engineer runs AI coding agents. The guide says the hard part moves from writing code to deciding who reads it, what proves it works and who answers for it after merge. It cites telemetry from Faros AI, an engineering analytics vendor, in which teams with high AI adoption merged 98 per cent more pull requests while review time rose 91 per cent.

• One instruction file, AGENTS.md or CLAUDE.md, is checked into each repository and reviewed like code, and rules that must hold go into hooks, automated checks and branch protection.
• A written review policy has the author review first, keeps changes small, sets review depth by risk and lets an AI agent read before a person does.
• Whoever merges the code owns it, and GitHub does not count the requester's approval of a pull request from its Copilot coding agent.
• Tickets name the files in scope, the acceptance criteria and the test that has to pass, and sprints are sized by the review hours available.

A team can check last month's agent-written pull requests for review wait, an approver other than the requester and a test that existed before the change, which is the guide's own first step. The guide is a staffing vendor's marketing page, and it says that the Faros figures are correlations from a company that sells analytics.

Derived by Working Surface from the article; more in the Playbook. Source line: Whoever merges the code owns it.

3 November 2026: teams working with AI agents are moving human review from each output to the plan, the shared instructions and the rules the agent follows beforehand, and naming in advance who answers for the result.