I see it all the time. Founders drowning in a sea of "ultimate guides" and "10-step frameworks" for building AI products. They’ve read every blog post, listened to every podcast, and yet their product is going nowhere. Stuck. Why? Because most of that advice is generic garbage. It’s written by people who have never actually built and sold an AI company. They’ve never had to answer to a board, to investors, to customers who are paying for a solution, not a science project. I have. I’ve been in the trenches, both as a founder with two exits and as an angel investor in over 200 companies, including some of the biggest names in AI like Anthropic, OpenAI, and Scale AI. I’m not here to give you another fluffy guide. I’m here to give you a reality check. A checklist that cuts through the noise and focuses on what actually matters. This is the stuff that will make or break your AI product.
The Counterintuitive Checklist for AI Product Success
1. Stop Obsessing Over the Model
I’m going to say something that might sound like heresy in Silicon Valley: your model is not your product. Not at first, anyway. I’ve seen so many first-time founders get hung up on building the most complex, groundbreaking, state-of-the-art model. They spend months, sometimes years, chasing a few extra percentage points of accuracy, while their competitors are out there shipping, learning, and iterating.
At MovieLaLa, my second company, we were building a movie recommendation engine. We could have spent a year trying to build a model that was marginally better than Netflix’s. Instead, we started with a simple collaborative filtering algorithm. It wasn’t perfect, but it was good enough. It allowed us to get a product into the hands of users within months. We learned more from that initial user feedback than we ever would have from a year of offline model tuning. The truth is, for most AI products, a fine-tuned open-source model is more than enough to get started. Your real differentiator is not the model itself, but the data you have, the user experience you build around it, and the problem you solve. Don’t tell me about your model’s architecture. Tell me about your data moat. Tell me about your distribution strategy. That’s what I care about as an investor.
Think about it this way: the early days of the internet weren't won by the companies with the most sophisticated algorithms. They were won by the companies that built the best user experiences and the strongest network effects. Google's PageRank was elegant, but it was the simplicity of the search box that won over the world. The same is true for AI. The winners won't be the ones with the fanciest models; they'll be the ones who build products that people actually want to use.
2. Your Data is a Mess. Fix It Now.
Everyone talks about data being the new oil. It’s a tired cliché, but it’s true. What they don’t tell you is that most data is a toxic sludge. It’s messy, incomplete, and full of biases. And if you feed your model toxic sludge, you’re going to get toxic sludge out.
I once had a portfolio company that was building an AI-powered tool for screening job candidates. They were so proud of their model’s accuracy. But when I dug into their data, I found that it was trained almost exclusively on resumes from a handful of elite universities. The model wasn’t just predicting good candidates; it was perpetuating a cycle of privilege. We had to go back to the drawing board. It was a painful and expensive lesson. Don’t make the same mistake. Before you write a single line of code for your model, you need to become a data janitor. It’s not glamorous work, but it’s the most important work you’ll do. You need to understand your data inside and out. Where does it come from? What are its biases? How are you cleaning and labeling it? You need to have a bulletproof process for data quality. If you don’t, you’re building your house on a foundation of sand.
This isn't just about bias. It's about noise, missing values, and a dozen other things that can trip up your model. I remember a time at RemoteTeam when we were building a feature to predict employee churn. The model was performing terribly, and we couldn't figure out why. It turned out that a single data entry error had created a whole set of phantom employees, throwing all of our calculations off. We spent weeks debugging the model, when the problem was in the data all along. That's why I tell my founders to invest in data infrastructure from day one. You need tools for data validation, data cleaning, and data versioning. You need to treat your data with the same care and rigor as you treat your code. It's the only way to build a robust and reliable AI product.
3. The "Human in the Loop" is Your Secret Weapon
There’s this fantasy in the AI world that you can build a fully autonomous system that just works, no humans required. That’s a myth. At least for now. The most successful AI products are not the ones that try to replace humans, but the ones that augment them. They build a seamless feedback loop between the model and a human expert.
At RemoteTeam, which was acquired by Gusto, we built a tool to help companies manage their remote employees. We used AI to automate a lot of the administrative tasks, but we always had a human in the loop to handle the edge cases and provide a personal touch. That human-in-the-loop component was our secret weapon. It allowed us to handle a much wider range of customer needs than our fully automated competitors. It also gave us a continuous stream of high-quality labeled data that we used to improve our models.
Think of your AI product as an apprentice. It’s going to make mistakes. It needs a human mentor to guide it, to correct it, and to teach it. Over time, the apprentice will get smarter and more autonomous. But in the beginning, you need that human in the loop. This isn't a sign of failure. It's a sign of intelligence. It shows that you understand the limitations of your technology and that you're focused on delivering a real solution to your customers. The best AI products are a collaboration between humans and machines, each playing to their strengths. The machine can process vast amounts of data and find patterns that a human would miss. The human can provide context, common sense, and empathy. Together, they can achieve things that neither could do alone.
4. Nail the User Experience. Seriously.
I’ve seen too many AI products that feel like they were designed by engineers for engineers. They’re a mess of confusing dashboards, technical jargon, and clunky interfaces. They might have the most powerful model in the world under the hood, but if the user can’t figure out how to use it, it’s worthless.
Your user doesn’t care about your model’s precision or recall. They care about whether your product solves their problem in an easy and intuitive way. They care about the user experience. This is where so many AI companies fail. They get so focused on the technology that they forget about the user. At MovieLaLa, we obsessed over the user experience. We wanted it to feel like magic. We spent as much time on the UI/UX as we did on the recommendation algorithm. We A/B tested everything, from the color of the buttons to the wording of the copy. That obsession with the user experience was a huge part of our success.
For AI products, good UX goes beyond just a pretty interface. It's about building trust and transparency. Your users need to understand what your AI is doing and why. They need to feel like they're in control. This is where things like explainable AI (XAI) come in. You don't need to show your users the raw output of your model, but you do need to give them some insight into how it's making its decisions. This could be as simple as highlighting the key features that influenced a prediction, or providing a natural language explanation of the model's reasoning. The goal is to demystify the AI and make it feel like a trusted partner, not a black box.
5. Your First 10 Customers are Your Co-Founders
You can’t build an AI product in a vacuum. You need to be in constant communication with your customers. Especially your first 10 customers. These are not just customers. They are your co-founders. They are the ones who will give you the honest, brutal feedback you need to build a great product. They are the ones who will help you find product-market fit.
You should be talking to them every single day. Get on the phone with them. Go visit them in their office. Watch them use your product. Understand their hopes, their fears, their workflows. I’ve seen founders who are afraid to talk to their customers. They’re afraid of hearing negative feedback. That’s a huge mistake. Negative feedback is a gift. It’s an opportunity to learn and to improve. The founders who embrace that feedback are the ones who succeed.
I have a rule for my portfolio companies: for the first six months, the CEO should be spending at least 50% of their time with customers. Not selling to them, but listening to them. I want to see customer interview notes, not just sales reports. I want to see that the product roadmap is being driven by customer needs, not by the latest AI hype. This is how you build a product that people love. It's how you build a company that lasts. So go out there and find your first 10 customers. Treat them like gold. And listen to them. They will tell you everything you need to know to build a billion-dollar company.
Building an AI product is hard. There’s no magic formula. But if you focus on these five things, you’ll be ahead of 99% of the competition. Stop chasing the hype and start focusing on what matters. Solve a real problem. Build a great product. And never, ever forget about the user. Now go build something amazing.
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.
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.
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.