Written by George Palmer-Bryant, Senior Software Engineer, Telana – originally posted on Linkedin here.

I recently wrapped up a high-stakes MVP for a major rail industry client, and it’s left me reflecting on how fundamentally the “developer workflow” has shifted. We didn’t just use AI as a better search engine; we shifted it left into the very core of our architectural and design processes.

Building a complex reporting dashboard in a compressed timeframe is always a challenge. Here are my three biggest takeaways on how AI helped us deliver significantly more value without sacrificing quality.

1. Shift AI Left: From Coding to Planning

The biggest mistake you can make with modern AI tools is only using them to write the code. On this project, we used AI for Solution Architecture and detailed planning before a single line of logic was committed.

By including clear context and actions into prompts, our agents were able to create structured plans to implement proven patterns. We identified potential bottlenecks before they became bugs. It’s about using the AI to validate the logic of your approach, not just the syntax.

2. The Power of “Directed Doubt”

One of the most effective techniques I found was telling the AI to doubt me. AI will both flatter you and take what you say as gospel. By telling the AI not to blindly agree with you, when you hit a wall you are much more likely to get an accurate answer as to where exactly you have gone wrong.

3. Navigating the “PoC Trap”

Our project started as a vibe coded PoC, and PoCs are often messy as the best of times. Our project was no exception, filled with technical debt like duplicate utility functions, inline SQL, and inconsistent data models.

The lesson learned here wasn’t just to “fix it all,” but to use AI to audit the debt. We used AI agents to identify dead code and highlight where our “Prototype Reality” (a mix of BigQuery and mock data) needed consolidation for production. It allowed us to move with the speed of a startup while maintaining a roadmap for enterprise-grade stability.

The Bottom Line:

AI didn’t replace the need for engineering, in fact, it highlighted the importance of being knowledgeable as an engineer on the age old principles like SOLID and DRY so we could identify and keep the good changes and cut the chaff. Ultimately, we were able to move faster and handle a project that might have traditionally required a much larger team or a much longer timeline.