Founder Insights and Lessons
Welcome back to Founder Mode!
I’ve been thinking a lot about what actually moves the needle when building an AI company. And here’s the thing: it’s rarely the stuff that sounds impressive.
The breakthroughs at Pretty Good AI didn’t come from better models or cleverer prompts. They came from fixing the boring parts. The parts that quietly slow everything down while everyone’s distracted by the shiny stuff.
This week, I want to share a few lessons straight from the trenches. These aren’t theories. These are changes we made because the old way stopped working.
The Redline Model
We used to treat every AI output like a creative draft. The model would generate a full document. A human would review the entire thing. Over and over.
It became a meat grinder. Slow, painful, expensive.
The fix was simple once we saw it. We stopped having AI rewrite everything from scratch. Instead, we start with a deterministic base document and have AI generate only the changes.
Now reviewers look at a redline diff instead of a fresh draft. Reviewing a few changes is ten times faster than reviewing an entire document.
If you’re building AI operations, this matters. Stop reviewing whole outputs. Start reviewing only the variance. Speed comes from reducing what humans have to touch.
The Trojan Horse Utility App
Selling a big enterprise platform is hard. Selling it cold is harder.
At PGA, we realized one of our customers’ biggest pains was something small and annoying. Manual data entry from insurance cards. Everyone hated it. No one expected it to be fixed.
Instead of hiding that fix inside our main product, we pulled it out. Made it a free standalone tool.
The goal wasn’t revenue. It was a distribution play.
When you solve one tiny, painful problem perfectly, people trust you. They share it. They come back. And suddenly, you have permission to talk about the bigger system.
Free doesn’t mean cheap. It means strategic.
Competitor Greed Is Your Entry Point
Competitors want big payments upfront. They ask before showing any value. Six-figure contracts. Long-term. Big promises.
That arrogance has become our advantage.
We offer zero-dollar pilots. This gets us meetings with enterprise customers we honestly shouldn’t be talking to yet. The contrast is sharp. One vendor asks for a check. We ask for a test.
When incumbents are friction-heavy, being frictionless lets you punch way above your weight.
If your competitors make it hard to say yes, make it easy. That alone can open doors you didn’t expect.
The New IP Battleground
As AI products mature, the real value often lives in the system prompt and knowledge configuration. That creates a new risk.
If you hand that configuration to a customer, what stops them from giving it to a cheaper competitor?
We learned quickly that standard copyright doesn’t protect this well. Trade secret status is also messy once client data is mixed in.
The solution for us was contractual. We updated our Terms of Use and NDAs to clearly define the configuration itself as proprietary work product.
If your prompts matter, protect them like code. Don’t assume the law will do it for you. As the saying goes... everyone has ideas. It's the execution that matters most!
Why This Matters
All of these lessons point to the same idea. AI success isn’t about being smarter. It’s about being operationally disciplined.
Winning teams remove review friction. They build trust with small wins. They lower barriers for new customers. They quietly protect their real assets.
That’s how we think about building PGA. Not as a demo machine, but as a system that works in the real world.
5 Key Takeaways
- Review less, not more. Redline diffs beat full document reviews every time.
- Use free tools strategically. A small utility can unlock massive distribution.
- Exploit competitor friction. Zero dollar pilots open doors greedy incumbents close.
- Protect your configurations. System prompts are intellectual property. Treat them that way.
- Optimize for operations. Real AI leverage lives in workflows, not models.
Final Thoughts
Building Pretty Good AI keeps reminding me that progress comes from humility. Admitting when something is broken. From fixing the boring parts others ignore.
The AI wave is moving fast. The companies that last won’t be the ones with the loudest claims. They’ll be the ones who quietly make life easier for customers and harder for competitors.
If you’re building right now, look for the redlines. Look for the friction. Look for the small, annoying problems everyone else has given up on.
That’s where the real leverage lives.
See you on Friday!
-kevin
Recent Social Posts
Recent Podcasts
2810 N Church St #87205, Wilmington, DE 19802
Unsubscribe · Preferences
