Incident management platform provider Rootly has announced that it is dropping its long-standing rule requiring small pull requests (PRs), a policy shift driven by the increasing role of AI agents in code generation. The company detailed its decision in a recent publication, explaining that the traditional metric of PR size no longer serves its intended purpose in the current development environment.

The rule, which had been a standard practice at Rootly, mandated that developers submit small, incremental code changes for review. The company now argues that with AI agents generating a significant portion of its codebase, the review process has changed, and the focus is shifting from measuring the sheer volume of changes to understanding their potential impact.

From Size to Blast Radius

Rootly’s new approach emphasizes a shift toward assessing the blast radius of a change, which refers to the potential scope of its impact on the system. This involves considering the number of services affected, user interactions, and dependencies, rather than raw line counts. The company states that this metric is more relevant for ensuring stability and security in an AI-assisted development workflow.

The move reflects a broader industry trend where AI code assistants are increasingly used to produce large volumes of code, making traditional review metrics like PR size less effective. In such an environment, a small PR can be risky if it touches a critical system, while a large PR might be safe if it only affects isolated components.

Feature Flags and Rollback Capability

To support this transition, Rootly is prioritizing the use of feature flags, which allow teams to deploy code that is not yet fully active, and robust rollback capabilities, enabling quick reversal if issues arise. These mechanisms provide a safety net that is more reliable than merely restricting the volume of code in a single review.

By using feature flags, developers can test changes in production without exposing them to all users, and rollback procedures ensure that any negative effects can be limited quickly. This approach is designed to reduce the risk associated with larger code changes generated by AI, while still maintaining code quality and system reliability.

The company’s decision is based on their observation that AI agents, which can now write and modify substantial parts of a codebase, have fundamentally changed the economics of code review. The cost of reviewing a large PR, in terms of time and cognitive load, is no longer directly proportional to its size, especially when automated tools can assist in the review.

Implications for Development Teams

Rootly’s announcement serves as an example for other engineering organizations that are grappling with the integration of AI into their development pipelines. The company’s move suggests that best practices for code review may need to be re-evaluated as AI tools become more common, with a greater emphasis on risk assessment and system-level thinking.

The shift also has implications for tooling and process design. Development teams may need to adopt new metrics and strategies, such as measuring the number of features affected or the time to rollback, rather than just PR size, to ensure effective oversight of AI-generated code. Rootly’s experience may prompt other companies to reconsider their own review guidelines.

Rootly has not announced a specific timeline for the full transition to its new review policy, but the company indicated that it is now actively applying this approach in its internal workflows. The company is also encouraging feedback from the developer community as it refines the method to better support AI-augmented software development.