I’m a builder. It’s in my blood. From my first startup, MovieLaLa, to RemoteTeam, I’ve always had this obsession with building things from scratch. It’s a trait that’s led to two successful exits and over 200 angel investments in companies like Anthropic, OpenAI, and Scale AI. But not long ago, this very instinct to build, to control every single layer of the stack, led me down a path that cost me a fortune. And I'm not just talking about money. I’m talking about time and focus—the only two things a founder can’t afford to waste.
It all kicked off with a simple, seductive idea: we could save a boatload of cash by building our own AI infrastructure. We were running a vertical SaaS company, and our monthly bill for third-party AI services was starting to look like a phone number. The math seemed like a slam dunk. At our scale, we were looking at a seven-figure annual spend on external APIs. My team and I huddled up, looked at the numbers, and said, "We can do this cheaper." We had the engineers. We had the ambition. What could possibly go wrong?
Famous last words.
The Siren Song of DIY
The decision to build our own AI infrastructure was based on a few assumptions that, in hindsight, were dangerously naive. We thought that by bringing everything in-house, we’d not only slash costs but also gain total control. We could fine-tune the models for our exact needs, cut latency, and build a proprietary asset that would be our secret weapon. On paper, it was a beautiful plan.
We dove in headfirst. We hired a squad of machine learning engineers and data scientists. We bought a mountain of servers. We burned months architecting a system that could handle our specific workload. The first signs were good. We had a working prototype, and the unit economics were looking sweet. We were on track to cut our AI-related expenses by over 60%. I was on top of the world. I thought we were geniuses.
But then, the universe decided to teach me a lesson in humility.
The Hidden Hell of Infrastructure
What we completely failed to price in were the hidden costs of keeping our own AI infrastructure alive. It wasn’t the servers or the initial development that bled us dry. It was the relentless, soul-crushing maintenance.
Here’s the real talk on what we didn’t see coming:
The People Problem: Hiring and keeping top-tier ML talent is a knife fight in a phone booth. We were in a constant bidding war for engineers who could build and maintain our systems. The salaries were insane, but the real cost was the distraction. My best people, who should have been obsessed with our core product, were now glorified IT support, babysitting a fleet of servers.
The Maintenance Treadmill: AI models aren’t a crockpot you can set and forget. They drift. They need constant monitoring, retraining, and fine-tuning. This isn’t a side project. It’s a full-time, all-consuming beast. We were so focused on the initial glory of the build that we totally underestimated the operational nightmare. Our shiny new infrastructure quickly became a black hole for engineering resources.
The Opportunity Cost: This was the dagger to the heart. Every single hour we poured into our internal AI platform was an hour we weren’t spending talking to customers, making our product better, and solving their actual problems. We were so in love with our clever technical solution that we completely lost the plot. We were a vertical SaaS company, not an infrastructure provider. Our obsession with saving money was, ironically, making our product worse.
The Epiphany: Back to First Principles
The wake-up call came during a board meeting. One of my investors, a guy who’s seen it all in SaaS, asked me a question so simple it was brutal: "Sahin, what business are you in?" I launched into my spiel about our AI infrastructure, our optimized models, our brilliant cost savings. He just held up a hand and cut me off. "You’re in the business of solving your customer’s workflow problems. Everything else is a distraction."
That hit me like a freight train. He was 100% right. We had fallen into the classic founder trap of getting high on our own technology. We were so fixated on the how that we had completely forgotten the why.
We made the gut-wrenching decision to kill our internal AI infrastructure. It was a bitter pill to swallow. We had torched millions of dollars and thousands of hours on the project. But it was the right call. We went back to using third-party AI services, and guess what? Our business started to breathe again.
My New Playbook for AI in SaaS
That expensive screw-up taught me a lesson I’ll never forget. In vertical SaaS, you don’t win by having the fanciest tech. You win by having the deepest, most intimate understanding of your customer’s world. You win by making their lives easier, more productive, and more profitable. Period.
Here’s the playbook I live by now when it comes to AI:
Obsess Over the Workflow, Not the Tech: Start by mapping out your customer’s entire day. Where are the bottlenecks? Where are the moments of frustration? Where can AI be a superpower? Don’t you dare start with the technology. Start with the human problem.
Embrace Serverless and Usage-Based Everything: The beauty of the modern AI stack is that you don’t have to build the plumbing. Services from OpenAI, Anthropic, and a dozen others give you world-class AI on tap. The rise of serverless AI and usage-based pricing is a gift. It means you can pay as you go and scale to the moon without ever having to rack a server. This is a superpower for startups. It lets you pour all your energy into your product, not your infrastructure.
Your Value is in the Application, Not the Pipes: Unless you’re in the business of selling picks and shovels, your competitive advantage will not come from building your own models. It will come from how you apply AI to a painful, specific business problem. It’s about the unique data you have, the proprietary workflow you’ve unlocked, and the ridiculously good user experience you provide. That’s your moat.
Looking back, I don’t regret the experience. It was an expensive education, but a necessary one. It forced me to get crystal clear on what actually matters: the customer. So here’s my advice to any founder out there in the SaaS world: check your ego at the door. Don’t build what you can buy. Focus on your unique genius, and let the experts handle the rest. Your customers will love you for it. And your bank account will too.
Frequently Asked Questions
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.
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 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.
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.