There is a popular narrative right now that AI-assisted development is like a "hot knife through ice cream"—effortless, magical, and instant. But as any engineer who has had to support "magic" in production knows: if you can't debug the magic, you don't own the stack.

I am an Inside Sales Engineer. My job is to bridge the gap between complex engineering and business value. Recently, I built the Liferay Docker Manager—a tool that allows our global team to spin up fully orchestrated Liferay environments (SSL, DBs, and Snapshots) entirely offline.

My velocity was incredible, but it wasn't "vibe coding." It was High-Speed Navigation.

The Reality of the "AI Experiment"

To make this tool "bulletproof" for my colleagues in the field, I had to apply three specific human disciplines that AI simply cannot replicate:

1. The Vision Filter (Scope Control) AI is a "yes-man." If you ask it to add a feature, it will—regardless of whether that feature adds technical debt or bloat. As the pilot, I had to act as the filter. I ensured every line of code fit the long-term roadmap: making Liferay accessible to non-technical stakeholders without needing a cloud connection.

2. The Debugging Tax (Ownership) The "magic" ends the moment a script fails on a colleague's machine 5,000 miles away. I spent hours hardening the logic AI suggested—refining how Docker handles volume persistence and ensuring the error messages were "human-readable." AI provided the boilerplate; I provided the "warts-and-all" hardening that makes it production-ready.

3. UX Intuition over AI Logic AI can’t "feel" a product experience. I focused on Convention over Configuration. For an SE visiting a customer site with no Wi-Fi, the tool needs to "just work." I spent my cognitive energy on the UX—designing a snapshot system that allows for fearless experimentation—while the AI handled the repetitive Bash and Python syntax.

The "Force Multiplier" for Sales Engineering

We didn't build this to replace our Cloud platform; we built it to democratise the local sandbox. By using AI to handle the "grunt work" of infrastructure orchestration, I was able to focus on the high-level architecture.

The result? A tool that allows any SE to walk into a "dead zone" and deliver a world-class Liferay demo, with the confidence that they can "reset to a known good state" at the click of a button.

Conclusion: Craftsmanship Matters

AI is a power-up, not a replacement for engineering craft. It allows us to move faster, but it requires us to work harder than ever on the vision and the underlying logic. It’s not magic—it’s discipline.