Kiran Kodithala is the founder and CEO of N2N Services, Inc.
Fifteen years ago, I stopped writing code. I had good reasons—or at least the reasons every founder tells himself. The company was growing. My job was strategy, customers and payroll. Building was for the builders.
Earlier this year, I went back into the code editor. My coach was my son.
Sit with that for a second. I’ve run a technology company for fifteen years. We serve more than 600 colleges and universities. And when it came time to learn AI-assisted development, the most qualified teacher I could find was a member of the generation we keep calling “too green.”
That reversal—the leader as student—is the whole story of this moment. Most executives haven’t accepted it yet.
The Half-Life Of Expertise Just Collapsed
Here’s the uncomfortable math. The experience curve that justified your seniority—and mine—was built on a stable toolchain. Learn the stack, compound the knowledge, cash in the authority. AI broke that model. Not slowly. In about eighteen months.
During an intensive internal sprint this year, I personally shipped hundreds of thousands of lines of production code with AI as co-author. Our small team shipped millions. That output would have been science fiction on our 2023 roadmap. But the number that actually changed how I lead was smaller: six.
Six junior developers—the kind résumé screens are built to filter out—fixed problems in days that veteran engineers hadn’t shipped in years. Not because they were smarter. Because they had no muscle memory to unlearn. They didn’t know the “right” way to do it, so they learned the new way instead.
Experience didn’t become worthless. But it stopped being a moat. In the age of AI, the compounding asset isn’t what you know. It’s how fast you’re willing to become a beginner again.
Why Leaders Can’t Delegate This One
The standard executive move is to delegate learning downward: fund a training budget, hire an AI lead, ask for a quarterly readout. I’m telling you that playbook fails, for one reason—you cannot make judgment calls about a medium you’ve never touched.
Every consequential AI decision on my desk this year—what to build, what to buy, who to trust, where the risks actually live—turned on texture I only had because I’d felt the tools myself. Where the models are brilliant. Where they confidently lie. What one developer can now do alone, and what still takes a team. You don’t get that from a briefing deck. You get it with your hands on the keyboard.
So what does learning look like when you’re the one signing the payroll?
Become a public student. I learned AI development from my son and said so—to my team, to our customers, to anyone who asked. The moment a leader admits ignorance out loud, the whole organization gets permission to learn instead of perform.
Judge output, not pedigree. Hand your least-credentialed people the same tools and the same hard problems you reserve for your seniors, then judge them only on what they ship. AI has collapsed the old barriers to building. Inclusion now means access.
Eat your own dog food. We recently mandated—mandated, not encouraged—that our team run its daily work on our own AI products. Not because the products are ready. Precisely because they aren’t. Nothing teaches faster than relying on what you built. Which brings me to the case study.
Twenty-Seven Days, On The Record
This summer, I put all three of those rules to a test I couldn’t fake. A three-person team—myself included—set out to build an enterprise platform from scratch, with AI agents doing most of the typing. Twenty-seven days later, we had shipped roughly 216,000 net lines of production code: a multi-tenant platform that turns one sentence into a running, governed application.
The number people quote back at me is the velocity—about 8,000 lines a day. The number that actually matters is different: 16 architecture decision records and 97 dated release notes. For every three lines of code, we wrote two lines of documentation. Fourteen automated domain reviewers gated every handoff. When the AI proposed a fix, it opened a real pull request—and a human merged it. When multi-tenant isolation bugs surfaced in real usage, we fixed them the day we found them and documented why.
Every leader I know would have delegated that sprint. I would have too, two years ago. But the lessons I use in every strategic decision now — where AI is brilliant, where it confidently lies, why velocity without governance is a demo and governed velocity is a business—none of them came from a status report. They came from the commit log with my name on it.
The Trip That Proved The Point
This month, I made my first trip to China and Taiwan, meeting hardware manufacturers for our Omnia Exec Assistant, N2N’s first AI-Edge product. I’m in my fifties. I’d never done hardware. I’d never done China. One partner spent nine hours walking us through their factories and every line of our specification. Every room I walked into, I was the least knowledgeable person there. That’s not a confession. That’s the strategy.
The Question That Matters
Leaders keep asking me some version of the question: “How do I future-proof my organization for AI?” Wrong question. Your organization will learn at exactly the speed you do. Teams watch what the leader does with their calendar, not what they say at the all-hands.
So the real question is simpler, and it’s the one I’d put to every executive reading this:
When was the last time you were bad at something in front of your team?
If you can’t remember, that’s your answer. In the age of AI, the leaders who win won’t be the ones with the most experience. They’ll be the ones who were willing to need a teacher—even if that teacher is your kid.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?