My ego nearly tanked my company.
It’s not a sentence you hear often in Silicon Valley, especially not from someone who has a couple of successful exits and a portfolio of over 200 angel investments. We’re taught to project success, to talk about the wins, the "up and to the right" charts. But today, I’m not going to do that. I’m going to tell you about my biggest leadership failure, a story of how my rush to implement AI brought one of my companies to the brink of collapse.
This is the vulnerable story of how I failed as an AI leader, and the hard lessons I learned clawing my way back. I’m sharing it because the hype around AI is deafening, and I see so many founders making the exact same near-fatal mistakes I did.
The Siren Song of the New
It was a few years back. The AI buzz was hitting a fever pitch. Every board meeting, every coffee chat with other founders, every tech headline was about AI. If you weren’t “doing AI,” you were a dinosaur. My inbox was a constant stream of pitches from startups promising to revolutionize everything with their new models. Even with my connections to teams at places like OpenAI and Anthropic, I felt the pressure. My ego got the best of me. I thought, “I’ve built successful companies. I’m an investor in the biggest AI labs. I need to be at the forefront of this.”
That was my first mistake. I became obsessed with the technology, not the problem it was supposed to solve.
At RemoteTeam, we were helping thousands of companies manage their global workforce. The operational complexity was immense. I saw AI as a silver bullet. I gathered my leadership team and announced a company-wide pivot: “We are becoming an AI-first company.” I painted a grand vision of a fully autonomous HR platform, one that could predict hiring needs, automate compliance, and personalize employee experiences, all with the click of a button.
It sounded incredible. The team was energized. I was a visionary leader. Or so I thought.
The Half-Million Dollar Bonfire
We dove in headfirst. I hired a team of expensive machine learning engineers, brilliant people who could talk for hours about transformer architectures and loss functions. We set up a massive data pipeline and started throwing money at GPU clusters. Our burn rate skyrocketed.
Our first major project was an “AI-powered onboarding assistant.” The goal was to have a chatbot that could answer any question a new hire might have, from “How do I set up my health insurance?” to “What’s the company policy on remote work stipends?”
We spent six months and burned through nearly $500,000 building it. The demos were slick. The chatbot could talk about our company history and even tell a few bad jokes. But when we put it in front of real users, it was a disaster.
The bot hallucinated answers. It confidently gave a new hire in the UK incorrect information about US tax forms. It couldn’t understand nuanced questions about our very specific reimbursement policies. Instead of reducing the burden on our HR team, it increased it. They were constantly apologizing for the bot’s mistakes and re-doing its work. It was worse than having no tool at all.
I had been so focused on building a "large language model" that I never stopped to ask if we even needed one. A well-structured FAQ page and a simple search bar would have solved 90% of the problem at 1% of the cost.
The team was demoralized. The engineers I hired were frustrated because they were trying to solve a poorly defined problem. The rest of the company saw the AI team as a black hole for money and resources, producing nothing of value. The tension was palpable. We were burning cash, our product was getting buggier because our focus was split, and our customers were getting angry. We were on a path to ruin.
The Humbling Intervention
The breaking point came during a team-wide demo. The AI assistant was meant to be the star of the show. I asked it a simple question about our vacation policy. It responded with a long, rambling, and completely wrong answer it seemed to have pulled from a random public company’s employee handbook.
There was a dead, awkward silence in the room.
Later that day, my co-founder and lead engineer, someone I had worked with for years, pulled me aside. He looked exhausted. “Sahin,” he said, “this isn’t working. The team is losing faith. I’m losing faith. We’re not an AI research lab. We’re a company that solves real problems for real people. And we’ve lost sight of that.”
It was a punch to the gut. But he was right. My vision wasn’t visionary; it was a distraction. My ego had written a check that our technology—and our strategy—couldn’t cash. I had fallen for the hype and dragged my entire company down with me. That night, I went home and had a long, hard look in the mirror. I had to admit to myself that I had failed.
The Painful, Necessary Pivot
The next morning, I called an all-hands meeting. I stood in front of the entire company and said, “I was wrong.”
I laid it all out: the wasted money, the flawed strategy, my personal failure of leadership. I apologized to the team for putting them through it. Then, I laid out the new plan. It wasn’t about abandoning AI. It was about being smart.
We shut down the general-purpose AI team. It was one of the hardest things I’ve ever had to do. We kept a few key people and embedded them directly into our product teams. Their new mandate was simple: don’t start with the tech, start with the customer. Find a small, specific, painful problem and see if a simple AI model could make it just 10% better.
Our first new project was completely different. We noticed our support team spent hours every week manually categorizing and routing support tickets. It was a repetitive, soul-crushing task. Could we automate it?
We didn’t need a massive language model. We used a simple classification model, trained on our own historical support data. It took two engineers three weeks to build and deploy. It wasn’t glamorous. It wouldn’t get us a headline in TechCrunch.
But it worked.
The model correctly categorized 95% of incoming tickets, saving the support team a collective 40 hours a week. That was a real, measurable impact. The team felt it. Our customers felt it, because response times got faster. It was a small win, but it was a real one. It rebuilt trust.
My New Rules for AI Leadership
That failure was my most valuable teacher. It reshaped my entire framework for building companies in the age of AI. Here are the rules I now live by:
Problem First, AI Second. Never start with “How can we use AI?” Start with “What is our most painful customer problem?” Only then should you ask if AI is the right tool for the job. Often, it’s not. A better process or a simpler bit of code is the answer.
Go for Boring, Incremental Wins. Forget the moonshots, at least at first. Find the boring, repetitive, and costly internal processes. Use simple AI to make them more efficient. These small wins build momentum and free up resources for bigger bets later.
Your Data Is Your Moat. The power of AI isn’t in the public models—everyone has access to those. The real advantage is your proprietary data. The model we built for support tickets worked because it was trained on our tickets, with our terminology and our customer issues. Focus on collecting and structuring a unique dataset that no one else has.
Embed, Don’t Isolate. Don’t create a separate “AI department” that is disconnected from the business. Embed your AI talent directly into your product and operational teams. They need to be sitting next to the people who feel the pain of the problems you’re trying to solve.
That turnaround didn’t just save the company; it made it stronger. We learned how to be surgical and strategic with technology. We stopped chasing hype and started delivering value.
The biggest threat in this new era isn’t the AI itself. It’s the hubris of the leaders implementing it. It’s easy to get seduced by the promise of god-like technology and lose sight of the fundamentals of building a business. My failure was a gift. It taught me that in the age of AI, the most important qualities for a leader are not vision or ambition, but humility and focus. Stay humble, and stay focused on the problem. The rest will follow.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.