We had the term sheet on the table. Five million dollars. For a founder, this is the moment you dream of. It’s validation. It’s fuel. It’s the starting gun for the race to build a massive company. I turned it down.
My team thought I was crazy. Our investors were confused. To be honest, for a second, I thought I was crazy too. But a nagging thought had been growing in my mind for months, a conclusion born from a painful, data-driven analysis that went against everything we’re told to do in Silicon Valley.
Everyone in the AI space is following the same tired advice. We decided to question it. After analyzing over 1,000 AI product failures, we found a shocking pattern that conventional wisdom completely misses. The data points to one uncomfortable truth about why most AI products never find traction.
The Siren Song of the AI Hype Cycle
Let’s be honest. The current AI landscape feels like a gold rush. You have incredible models from my friends at OpenAI and Anthropic, and the pressure is immense. The playbook is clear: raise a huge seed round, hire a team of expensive ML engineers, build a technically impressive model, and then figure out how to shoehorn it into a problem. The focus is on the what—the technology—not the who and the why.
This is the advice I’ve heard in countless boardrooms and from VCs on Sand Hill Road. It’s the path that leads to glowing TechCrunch articles and a billion-dollar valuation. It’s also, as our research showed, the path to ruin for the vast majority of AI startups. I've seen it firsthand in my own portfolio of over 200 angel investments. I've had to write off checks to brilliant teams who built technological marvels that nobody wanted to buy.
We were on that exact path. We had a brilliant team, a powerful idea, and a demo that investors loved. The $5 million was supposed to be our rocket fuel. But as we prepared for the fundraise, I started a side project. I wanted to understand why so many promising AI companies I had seen or invested in were quietly fizzling out after raising big rounds. What was going wrong? Was it the team? The market timing? Or something more fundamental?
The 1,000-Failure Study
I tasked a small, scrappy team with a simple goal: find 1,000 AI products that had been funded in the last five years and had failed to achieve any meaningful product-market fit. We weren't looking at spectacular flameouts, but the silent majority that just… faded away. We dug through post-mortems, scraped LinkedIn for former employees, and spent weeks interviewing founders who were willing to talk. It was emotional archaeology.
It was a depressing process. We heard stories of incredible technical talent burning out trying to find a use case for their solution. We saw pitch decks that promised to revolutionize entire industries, followed by balance sheets that showed zero revenue. But amid the wreckage, a pattern emerged. It was so consistent and so contrary to the popular narrative that I almost didn’t believe it.
The failed companies all chased the technology first. They started with the question, “What can we build with this cool new AI?” They spent millions on R&D before they had a single paying customer who was desperate for their solution. They built complex, elegant models to solve problems that users didn’t really have, or that could be solved with a much simpler, non-AI solution. One founder told us, “We spent two years and $3 million building an AI to predict customer churn. It turned out our clients just needed a better onboarding process.” That one quote summed up hundreds of failures.
Conversely, when we looked at the breakout AI successes, we saw the opposite pattern. Almost every single one of them started with a painful, specific, and often boring problem. They solved it manually at first—what we called the “Wizard of Oz” phase. They acted like the AI, using humans to deliver the service, to understand the workflow, the edge cases, and the true user need. They only started layering in AI piece by piece to automate the most repetitive, well-understood parts of the process. Think of it like building a self-driving car. You don't start by building a fully autonomous vehicle. You start with cruise control, then lane assist, then automatic braking. You automate the simple, solved problems first.
The uncomfortable truth was this: Successful AI products don't start with AI. They start with a service that is so valuable, people would pay for it even if it were powered by a human in a spreadsheet.
AI is a scaling mechanism, not a product in itself.
A Painful Realization
This hit me like a ton of bricks. Our entire strategy was built on the flawed model. We were about to take $5 million and pour it into building a complex AI core before we had truly validated the desperation of the user need. That money would have come with expectations. We’d have to hire that big ML team. We’d have a high burn rate. We’d be locked into our initial vision, a vision that our own research now told us was likely wrong.
I remembered a lesson from my first company, MovieLaLa. We gained traction not by having the most complex recommendation engine, but by manually curating movie lists and focusing on the community aspect. The tech came later, after we knew exactly what people wanted. At RemoteTeam, which was acquired by Gusto, we didn't start with an AI to manage global payroll. We started by manually helping a few companies with the painful, messy process of paying international contractors. We were the software. We built the actual software to automate the things we were already doing, one painful step at a time.
Taking that $5 million would have been a trap. It would have forced us to build a solution in search of a problem. It would have been a betrayal of the most important lesson I’ve learned in my career as a founder and investor. It’s a lesson I try to instill in every founder I advise: fall in love with the problem, not your solution.
So, I walked away. I called the lead investor and told him we were pulling out. There was a long, stunned silence on the other end of the line. Then he said, “Sahin, in ten years of doing this, I’ve never had a founder turn down a term sheet at this stage.” I told him I’d rather return the money we hadn't yet taken than waste the money we would have.
The New Vision: Manual First, AI Second
We’re starting over. Not from scratch, but with a completely new philosophy. Our new vision is built on a few core principles:
Obsess over a tiny, painful problem. We’re not trying to build a platform to change the world. We’re trying to find a workflow that is so broken and so costly for a specific group of people that they would happily pay for a solution tomorrow. We are interviewing dozens of potential customers a week, not to pitch them, but to listen. We are looking for the pain in their voice.
Deliver a “Wizard of Oz” service. We will be the AI. We will manually deliver the solution to our first ten customers. We will live their pain, document every step, and understand the problem better than anyone else on the planet. This means our first hires might be customer support agents, not data scientists. It means our product roadmap is a Google Doc, not a Jira board.
Automate reluctantly. Only when a step in our manual process is so well-understood and repetitive that a human is bored doing it will we replace it with a piece of code. We will gradually and carefully layer in AI to make our manual service more efficient, not to replace the human insight that makes it valuable. The goal is to augment our human experts, to give them superpowers, not to replace them. This is a critical distinction that most AI companies get wrong. They see AI as a cost-saving measure, a way to get rid of expensive humans. We see it as a quality-enhancing measure, a way to free up our experts to focus on the highest-value, most complex parts of the problem.
A New Kind of Team
This new vision also requires a new kind of team. We’re not looking for the stereotypical Silicon Valley engineer who wants to work on the latest and greatest tech. We’re looking for people who are obsessed with solving customer problems. We’re looking for people who are willing to get their hands dirty, who are comfortable with ambiguity, and who are excited about the prospect of building a business from the ground up.
Our first hire after making this pivot was not an ML engineer, but a former management consultant who is an expert in process mapping. Her job is to document every single step of our manual service, to identify the bottlenecks, and to help us figure out where technology can make the biggest impact. We’re building a team of problem-solvers, not just technologists.
This is a harder path. It’s less glamorous. It doesn’t generate the same kind of hype. But it’s a path that we believe is far more likely to lead to a sustainable, profitable business. It’s a path that is built on a foundation of real value, not just speculation.
The Future is Boring
Everyone in Silicon Valley is chasing the sexy, world-changing ideas. But the biggest opportunities are often in the boring, unsexy, and overlooked corners of the market. The problems that are so painful and so ingrained in the way people work that they’ve just accepted them as a cost of doing business.
These are the problems that are ripe for disruption. And they are the problems that are best solved with the manual-first, AI-second approach. By starting with a deep understanding of the problem and by building a solution that is tailored to the specific needs of the customer, we can create a product that is not just a nice-to-have, but a must-have.
Turning down that $5 million was a bet. It was a bet that we could build a better company by ignoring the conventional wisdom. It was a bet that by focusing on the customer first, we could build a product that would not just survive, but thrive in the crowded AI landscape. It’s a bet I’m willing to make.
The next generation of great AI companies won’t be built by the teams that are the best at building AI. They will be built by the teams that are the best at solving problems. And that is exactly what we intend to do.
Frequently Asked Questions
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.
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.
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.
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.