I’ve been in Silicon Valley for a while. Long enough to start a few companies, sell two of them (you might know RemoteTeam, which was acquired by Gusto, or MovieLaLa, which went to Gfycat), and write checks to over 200 startups. I’ve been lucky to get in early on some incredible companies like Anthropic, OpenAI, and Scale AI.
I’ve seen what makes companies fly and what makes them crash and burn. And the crashes that hurt the most are the ones that should have been rockets. Brilliant people, incredible tech, and a product that just… sits there. No one uses it. No one buys it.
We were about to become one of them.
The AI Roadmap to Nowhere
There’s a playbook for building AI products that everyone seems to follow. It’s almost a cliché at this point.
- You find some hot new model.
- You build a slick demo that makes the tech look like magic.
- You create a roadmap that’s all about making the tech even more magical.
- Then you go looking for someone, anyone, who will pay for your magic trick.
That was us. We had a team of absolute rockstar engineers, a roadmap packed with impressive-sounding milestones, and a product that was, technically speaking, a marvel. We were so proud. We thought users would be lining up around the block.
Then we talked to a real user. And the whole thing fell apart.
The User Interview That Felt Like a Punch to the Gut
It was a Tuesday. I almost canceled the call. The guy was a manager at a small logistics company. Not exactly our “ideal customer.” But we had the time slot, so I figured, what the hell.
For the first 20 minutes, it was perfect. We ran through the demo. He nodded. He said things like, “Oh, that’s neat.” and “I can see how that could be useful.” The classic signs of someone being polite, not someone whose problem you’re solving.
Then I asked the question that killed us: “So, how would you actually use this in your day-to-day?”
Dead silence. Not just a pause. It was the kind of silence where you can hear the air conditioner humming. He just stared at the screen.
Finally, he started talking. But he wasn’t talking about our product. He was talking about his problems. The nightmare of scheduling his truck drivers. The endless, soul-crushing email chains and spreadsheets he was buried in. His biggest problem wasn’t that he lacked some futuristic AI. His biggest problem was that he was drowning in basic, boring operational chaos.
Our product didn’t solve a single one of his real problems. Not one. It was a beautiful, expensive solution for a problem nobody had. In that moment, I realized we’d spent six months building a spaceship for a guy who just needed a better shovel.
I walked out of that meeting feeling like I’d been punched in the stomach. Our 18-month roadmap, filled with buzzwords like ‘multi-modal inputs’ and ‘generative adversarial networks,’ was a complete fantasy. Not a single line item on it would have helped that guy with his truck drivers.
That afternoon, we threw the entire roadmap in the trash.
The Dirty Secret of Most AI Products
That interview sent me down a rabbit hole. I started talking to dozens of other founders. I dug into the data on over 1,000 failed AI products. A pattern emerged. A really uncomfortable one that no one in the Valley likes to talk about.
Most AI products don’t fail because the tech is bad. They fail because they’re built on fantasies. They’re built to impress other engineers, not to solve a customer’s problem.
We fall in love with our own cleverness. We get so wrapped up in what the technology can do that we never stop to ask what it should do. We build things that are technically amazing but practically useless. And then we act surprised when nobody opens their wallet.
My own history proves this out. When we built MovieLaLa, we didn’t start with some fancy algorithm. We started with a dead-simple problem: movie studios needed to sell tickets to teenagers, and teenagers were on their phones. We just connected the dots. It worked. Gfycat bought it.
It was the same story with RemoteTeam. The world was going remote, and companies had no idea how to hire or pay people in other countries. It was a huge, expensive headache. We built a platform that made it simple. It wasn’t about having the smartest AI for payroll; it was about solving a real, painful business problem. That’s why a $10 billion company like Gusto was interested. We solved a problem they had.
A Better Roadmap
So what’s the right way? It means flipping the script completely. It means forgetting about the tech for a minute and getting obsessed with your user.
Here’s the new playbook. It’s not as sexy, but it actually works.
- Find the screaming pain. Don’t start with tech. Start with a person. A real person with a real, expensive, hair-on-fire problem. Go live in their world. Watch them work. Find the thing that makes them curse at their computer screen. That’s where you start.
- Build the dumbest possible solution. Seriously. Once you find that pain, build the most basic, simple, un-fancy thing you can to make it go away. If a simple form and a database can solve 80% of the problem, do that first. Get people using it.
- Then, and only then, get smart. Once you have a product that people are actually using and paying for, that’s when you look for ways to sprinkle in the AI. Can you automate something? Can you personalize the experience? Can you predict something useful? The AI should be the turbocharger, not the engine.
This is way harder. It’s less glamorous. You don’t get to post exciting demos on Twitter every week. But it’s how you build a real business, not just a cool science project.
I’m not saying tech doesn’t matter. Of course it does. But it’s a tool. It’s a means to an end. The end is always—always—solving a real problem for a real human. If you lose sight of that, you’re just building another beautiful headstone for the AI product graveyard.
Throwing out our roadmap was the best thing we ever did. It was a painful, humbling, and necessary dose of reality. It forced us to stop admiring the problem and start actually solving it.
Frequently Asked Questions
How often is this guide updated?
I revisit and update my guides regularly as I learn new things and as the market evolves. The core principles tend to stay stable, but specific tactics and tools get refreshed based on what's working right now.
Who is this guide designed for?
This guide is written for founders and operators who want practical, actionable advice rather than theoretical frameworks. Whether you're just starting out or scaling an existing business, the principles here apply across stages.
Is this guide based on real experience?
Every recommendation in this guide comes from direct experience, either from building and selling my own companies, or from patterns I've observed across 200+ angel investments. I don't write about things I haven't personally tested.