Steffen Froehlich uses AI to review code written by other AI systems — catching bugs that slow down development

Froehlich, an engineer at DocuSign, described using AI (specifically large language models that understand code) to review and critique code that was generated by other AI coding assistants. Instead of relying solely on human code review, he's treating AI as a second layer of quality control.

The approach works by having one AI system explain potential problems in AI-generated code — like performance bottlenecks or logical errors — before a human engineer even looks at it. This catches obvious mistakes earlier in the review cycle, reducing the back-and-forth that slows down shipping.

For engineering teams using AI coding tools, this means code reviews can move faster because AI pre-screens for common flaws, freeing human reviewers to focus on architectural decisions and business logic. Teams shipping features weekly instead of bi-weekly could see a measurable difference.