AI has become woven into nearly every aspect of our lives. Need a travel itinerary? ChatGPT can handle it. Want a silly photo for your friends? AI is there, too. Yet, the term “AI” itself feels more like a marketing buzzword than a technical description. “LLM” (large language model) is more accurate, but it tends to make people’s eyes glaze over.

For developers, AI is now an essential tool—sometimes for better, sometimes for worse. Entering the field of software development today is more challenging than ever. While AI makes it easier to complete tasks, truly understanding the underlying concepts and best practices will still be a major hurdle.

Documentation is easier to create, but it still needs to be validated and understood—just like everything else in technology.

AI will improve

AI is good, but not perfect. I’ve been using GitHub Copilot Pro and find it worth the subscription. The assistant in Visual Studio and VS Code feels like pair programming with someone who can help when you’re stuck. The new agent mode (currently in preview) is a game changer. When working with codebases that have strong patterns and clear data flows, it really shines. I can let the agent handle routine tasks while I focus on more creative work—like writing this blog.

Areas where it still needs improvement:

  • Sometimes breaks the build
  • Generated tests do not always pass
  • Missing or incomplete functionality
  • Requires manual verification and QA

I don’t think we’re at the point where AI can run an entire codebase on its own—nor do I think we ever truly will—but who knows?

Project Management will need to keep up

After using GitHub Copilot Pro and its agent mode for a while, I’ve realized that effective project management is more important than ever. Directing AI efficiently requires a solid understanding of the project’s structure and goals. Laying a strong foundation is crucial for high-quality results. Some tasks will involve cleaning up technical debt or preparing for future improvements. Project management may struggle to balance business needs with the technical requirements for long-term success. Coding is speeding up, and the bottleneck is no longer the code itself—it’s about what we’re building, how we’re building it, and the timeframe. These questions will become even more important.

The Core is still the Core

AI won’t replace the core concepts that are crucial to software development—TDD (test-driven development), separation of concerns, N-tier applications, integration/unit tests, and more. It will improve test coverage and help scaffold the core functions needed for these concepts. Adding code standards and best practices to AI workflows will only enhance the software development lifecycle. But while AI makes it easier to complete tasks, truly understanding the underlying concepts and best practices remains a major hurdle.

Keeping everyone aligned and ensuring that essential standards are met is critical for long-term success. I’m still learning how to leverage AI more effectively—continuous improvement is a constant—in life and in Information Technology.

The only constant is change. —Heraclitus

In closing

Maybe AI wrote this. Maybe I did. Does it matter? Ultimately, what matters is the quality and impact of the work.