I spent decades mastering software architecture, only to discover that 70% of my hard-won expertise could now be delivered by AI in seconds. This should have terrified me. Instead, it set me free.
Let me back up. I started my career as a software engineer and quickly progressed through the ranks to architect. Around the start of this century, Microsoft launched their Microsoft Certified Architects program—an exclusive certification requiring peer review of real-world architecture case studies. I flew to Singapore for my board review and became one of only 40 architects worldwide to earn that certification before Microsoft sunset the program due to costs. I co-authored a book on Professional UML with Visio, back when UML was the graphical standard for designing architecture.
I’m telling you this not to wave credentials, but to establish something important: I lived and breathed architecture for years. And through all that experience, I discovered something fundamental.
READ PART 1: Why your AI coding agent needs more than a plan: Lessons from the trenches
The 30/70 Rule
At least 30% of architecture work is the art of asking the right questions. This came naturally to me—maybe it was intellectual curiosity, maybe just common sense. But I watched talented engineers fail because they never asked those critical questions. You ask them of customers. You ask them of yourself. You ask them of experienced engineers. Through these questions, you arrive at the right architectural decisions.
But here’s the thing: common sense isn’t always that common.
The other 70%? That required immense industry experience and knowledge. Constantly reading. Learning from your own mistakes—or better yet, someone else’s. If you’re sitting high enough in a big company like Google, you get exposed to internal mistakes and successes through a variety of different teams. But basically, you’re constantly sucking in a tremendous amount of information so you have that contextual awareness.
This knowledge was my competitive advantage. But unlike fine wine, technical knowledge has an expiration date.
The Decay Problem
As I moved from pure architecture into building businesses—creating category-defining companies like Wrike (which helped create collaborative work management alongside Asana and Monday) and now Zencoder (building AI agents that automate the entire software development lifecycle)—I drifted further from day-to-day technical decisions. Technologies evolved. Patterns shifted. Best practices transformed.
My bottled-up knowledge was turning to vinegar.
Recently, I needed to dive deep into some hard architectural decisions. What happened next changed everything I believed about expertise.
The AI Multiplier Effect
My mind was blown by what I could accomplish with AI assistance. That 70% of industry knowledge I’d spent years accumulating? AI could deliver it instantly, accurately, and comprehensively.
Here’s my new process: When facing an architectural question, I immediately fire up at least three AI queries. If the question requires analyzing an existing codebase that I’m not familiar with, I open it in Zencoder and fire a bunch of parallel queries to the agent. In parallel, I ask GPT5 or Opus for the industry best practices perspective. Then, while it’s working, I copy-paste the industry question to “deeper” public research agents – GPT-5 Pro and Research Mode in Claude Opus. By the time I launch my last query, the first one already has a response for me, and it’s time to dive into the rabbit hole.
It’s like being a detective with a team of tireless investigators. The initial response gives me threads to follow. Like any good detective, I pull on those threads—they lead to more threads. It’s a back-and-forth process, diving deeper into promising directions. By the time the research agents return with their full write-ups, I’ve already formed initial opinions that I can validate or revise against their comprehensive analyses.
What happens then is I can act as a real architect again, even after years away from daily practice. My prior wisdom and battle scars combine with AI’s up-to-date knowledge to create something neither could achieve alone.
The New Architecture Equation
The equation has fundamentally changed:
- Before: 30% asking right questions + 70% accumulated knowledge = Good architecture
- Now: 30% asking right questions + AI-delivered knowledge + Human judgment = Exceptional architecture
This isn’t about AI replacing architects. It’s about AI completing us. The architects who will thrive aren’t those with the most memorized patterns or the deepest technical knowledge. They’re the ones who ask the best questions, who know which threads to pull, who can synthesize AI-delivered insights with human judgment and experience.
The Paradox of Progress
Everyone fears AI will replace architects. After building multiple category-defining companies, I’ve discovered the opposite truth: AI doesn’t diminish the value of experience—it amplifies it.
My “expired” knowledge isn’t worthless. Those scars, those lessons learned, that intuition about what questions to ask—these become more valuable, not less, when paired with AI’s infinite capacity for current information.
I was one of 40 Microsoft Certified Architects in the world. Today, I’d lose a technical knowledge contest to any junior developer with ChatGPT. Yet I’m building better systems now than I ever did at my peak.
That’s not a contradiction. It’s evolution.
The 30% that was always the art of architecture—that’s now where the entire game is won. And if you’ve spent years honing that art, if you have the scars and wisdom from real-world systems, then AI isn’t your replacement.
It’s your superpower.
