In a recent LinkedIn post, Melissa Perri discusses a foundational yet often overlooked decision that significantly improved her team’s website development process, particularly in the context of using AI tools like Claude Code. Perri highlights how integrating design and content workflows into Git repositories, mirroring the practices of software developers, led to substantial gains in efficiency and clarity.
Perri explains the common pain points of traditional handoffs between design, content, and engineering teams, which often involve cumbersome file sharing and a lack of clear version history. To combat this, her team implemented a novel approach:
“We put design and content into the same Git repositories as the code. Even our images and logos live in source control now.”
This decision, though seemingly unglamorous, required the entire team to acquire new skills. Perri herself, despite having a GitHub account for years, had to actively engage with the platform, making multiple commits daily. She notes the learning curve was manageable:
“We each had to learn it, or in my case relearn it, and it only took a few days to feel natural.”
The Power of a Single Source of Truth
Perri emphasizes the profound impact of having a unified, timestamped record for all changes. This contrasts sharply with fragmented systems where multiple versions and sources of truth exist. As Perri points out, this structured approach is particularly crucial when working with AI.
AI Integration and Enhanced Development Workflows
A key takeaway from Perri’s post is how version control facilitates more effective AI collaboration. She explains that AI models like Claude Code can leverage Git’s capabilities to understand changes more precisely.
Streamlining AI Execution
According to Perri, AI can compare existing code with new commits, enabling it to make targeted modifications rather than starting from scratch. This clarity in input significantly reduces the effort required to prepare assets for AI execution.
“When the inputs are this clear, ‘ready for development’ and ‘ready for the AI to execute’ become almost the same thing.”
Furthermore, Perri details how her team established rigorous definitions of “done” for development tasks, including meeting requirements, implementing error handling, passing tests, and checking edge cases. They also focused on keeping changes small and contained, with automated performance checks to ensure pages load quickly and maintain search rankings.
Compounding Benefits and Cultural Shift
The compounding nature of these improvements is a significant benefit, as Perri describes. Lessons learned and reusable patterns are now embedded within the codebase, accessible for future projects and AI sessions. This allows both human team members and AI to build upon existing work rather than reinventing the wheel.
The shift also fostered a positive cultural change, as highlighted by Perri’s engineering director. Rather than feeling like their skills were becoming obsolete, the team experienced the best practices of software development—version control, traceability, and a clear record of changes—permeating throughout the entire team. This, Perri concludes, is fundamental to making AI-assisted development truly effective.
📝 About This Content
This article is based on insights shared by Melissa Perri on LinkedIn.
📅 Originally posted on July 31, 2026 | View original post on LinkedIn →