How to Start a SaaS Company

I'm writing this while my Claude Code agent builds out a new piece of our SaaS in the background. Feels like the right headspace to talk about how you actually start one of these things. So here's the real playbook, not the guru version.

Step 1: Find a real problem

Before you write a single line of code, go talk to your customers. Actual humans with actual pain. And take good notes, thorough ones. This is the most important and most skipped step.

Don't stop at the surface. Keep asking why. Say the problem is social media management. Why is that a pain? "It takes forever." Why else? "The agencies are expensive." And? "None of them sound like me." Now you're getting somewhere. Each "why" peels back a layer, and the layers are where the product actually lives.

These days you can dump all those notes into your AI tool of choice and it'll hand you a rough spec in minutes. I built mine in the early days of AI and didn't have that luxury, so trust me, use it. But the spec is only as good as your notes, and the notes only come from actually listening.

A quick word on pre-selling: I don't buy it

A lot of people will tell you to pre-sell before you build. I disagree, at least for most people. If you're a seasoned software vet who can conjure the thing on demand, sure, go ahead. But if you're just starting out, pre-selling puts the cart miles ahead of the horse. You need to go through the actual process of building something first, because that process is where you learn what you're even selling.

Step 2: Build an ugly MVP

Your first version should be duct tape and chewing gum. I mean that literally.

My first real product was something we called the ASP form. A business owner uploads a photo to a form, ChatGPT writes a caption in their voice, and it posts to all their social profiles. Simple, useful. And the first version was an absolute horror show under the hood: a GoHighLevel form, n8n middleware, a ChatGPT API call, and then a handoff to a Zapier endpoint, because the social posting tool I was using back then (OneUp) didn't even have a public API.

It was ugly. It was disgusting. And it worked. That's the entire point of an MVP. Use cheap tools, skip the bells and whistles, let it be held together with tape, but make sure it actually does the job and solves the problem. Working and ugly beats beautiful and imaginary, every single time.

Step 3: Price it low, or give it away

Now put it in front of people, cheap or free. And here's why free is fine at this stage: you're not optimizing for revenue yet. You're after the single most valuable thing a young startup can get, honest feedback from real users.

Every user who touches your ugly little product is telling you what to build next. That feedback is worth far more than the few bucks you'd charge for v1. This isn't a trick to get something out of people, it's the opposite. You give them real value, and in return they tell you exactly how to make it better.

Step 4: Iterate faster than everyone else

This is the one that actually matters. As a startup, speed is your only real advantage. The faster you iterate, the better your product gets, and the harder you are to catch.

Let me show you our own embarrassing evolution, because it proves the point. We started out white-labeling GoHighLevel, with a backend that was pure duct tape: Zapier, GHL, and me hand-building every workflow one at a time. Then I learned what a database actually was. Then we rebuilt the backend the right way and made it multi-tenant, which just means the whole thing got productized: instead of me babysitting 30 separate workflows, every user hits one shared database and the same logic serves all of them. Then, finally, I used Claude Code to build our own custom product from the ground up.

From glued-together GHL to a real SaaS platform. The only reason we made that jump is that we obsessed over iterating fast at every single stage. We improved faster than the people around us. That's the whole game.

About AI: it's leverage, not magic

Since AI is baked into all of this now, one hard truth. AI is not magic, and it will not replace good engineering judgment. I've watched it make genuinely horrible mistakes, and if you're not paying attention to catch them, you're in for a world of pain.

What AI does do is let you iterate at a speed that used to be impossible. I built our SaaS products in about two and a half months of 16-hour days. Without AI, that's easily two or three years of work. I'm as bullish on it as anyone, just be smart. Drive it, don't trust it blindly.

And here's the beautiful part of getting to a real, custom-coded product: adding a new customer barely costs you anything, and pulling in new users stops being a heroic effort. The economics finally start working for you.

One more thing: get a great mentor

I would not be where I am without the mentors I've had. Full stop. Find people who've already walked the road you're on, and actually listen to them. It's one of the highest-leverage things on this entire list, and it doesn't cost you a single line of code.

The whole thing in four lines

1. Talk to your users and find a real problem.
2. Build something that works, even if it's ugly duct tape.
3. Iterate faster than your competition.
4. Then go talk to your users again, and keep the loop running.

That's how you start a SaaS company. Not in a weekend, but a lot faster than you'd think if you stay obsessed with the loop.

Building something and want it built right?

Real architecture, real iteration speed, and judgment where it counts. That's what we do at HyppoAI.

Build it right