As technology consultants, staying at the forefront of innovation isn't just a competitive advantage—it's a necessity. Over the past year, our development team has strategically integrated AI tools into our workflow, dramatically improving our efficiency while maintaining our high standards for code quality. Here's how we're doing it.
Although we’re constantly evaluating and experimenting with new tools as they come along, we've narrowed our focus on using just a couple of tools that integrate seamlessly with our existing development environment:
Claude Code has become an invaluable part of our stack, allowing our developers to delegate coding tasks directly from their terminals. This integration means less context switching and more focused development time.
GitHub Co-pilot Agent Mode extends our capabilities within the GitHub ecosystem, helping with everything from code generation to debugging and refactoring. The ability to have an AI assistant that understands the full context of our repositories has been transformative.
We've learned that AI tools aren't magical solutions that work perfectly out of the box. Our approach to "Vibe Coding" (letting AI generate substantial portions of our codebase) comes with carefully designed guardrails:
Perhaps the most critical lesson we've learned is that the quality of AI output directly correlates with the quality of our Product Requirements Documents (PRDs). Vague descriptions produce unpredictable results, while detailed, well-structured PRDs yield code that matches our expectations.
Making effective use of Model Context Protocol (MCP) is starting to transform what we can do with AI, well beyond guided coding. One place where it’s clearly helped is that we’ve started integrating Supabase with MCP, for things like:
Some of this has been possible with straight AI tools for a while (without MCP in the mix), but as we integrate more pieces of the puzzle, there’s a compounding benefit. Fold in Jira, and now it can access ticket details, requirements, and acceptance criteria directly. Add Figma integration and the AI now has access to design specifications too. It’s not just an efficiency booster, but as we round out the picture about what we’re trying to accomplish for the AI tools, the better it gets at helping us accomplish our goals.
So, as we incorporate these tools across disciplines, we’re finding it allows everyone on the team—product managers, designers and developers—to focus more on making sure we’re solving the right problems for the business and its customers.
Our journey with AI development tools has taught us that the key to success lies not in blindly embracing automation, but in thoughtfully integrating it within a framework of best practices, clear documentation, and human oversight.
By providing structure and guidance to AI tools, we're able to harness their power while ensuring the code produced meets our standards for maintainability, extensibility, and performance. The result is faster development cycles without compromising on quality—a win for both our team and our clients.