Code review, in the agentic era.

In the agentic SDLC, agents write, test and open pull requests faster than people can read them, so review has become the bottleneck. The teams handling it best have stopped reviewing every line the same way.

• Review by risk, not by author. Low-risk changes merge on automated checks. People look at architecture, direction and risky code.
• Machines check first. The agent verifies its own work, then CI and AI reviewers run before a person is asked.
• Evidence before review. Every PR states its intent and shows proof it runs: tests, output, small diffs.
• A person owns the merge. AI feedback is a signal. An agent’s approval never counts as a human’s.

Reading list

• How the Claude Code team does code review — claude.dev, Oct 2026. Layered review: self-checks, CI, multi-agent review, then risk-based human approval. Merged PRs doubled, reverts down 40%.
• Agentic Code Review — Addy Osmani, Jun 2026. Scale review depth to risk, demand evidence before reviewing, read the tests first, and treat AI feedback as a sensor, not a verdict.
• Agent pull requests are everywhere. Here’s how to review them. — The GitHub Blog, May 2026. A practical reviewer’s guide to agent PRs, and the red flags to look for, like duplicating existing code.
• Orchestrating AI Code Review at scale — Cloudflare, Apr 2026. Up to seven specialised reviewer agents on every merge request, run in CI.
• What is happening with code reviews? — The Pragmatic Engineer, 2026. Agents open more and bigger PRs than people ever did. How teams are coping.

Linescape: visualize your team’s flow, spot the bottlenecks at a glance.