Give a small boy a hammer and he will find that everything he encounters needs pounding.
— Abraham Kaplan
AI is the latest hammer; founders are the small boy. We’re all busy perfecting our tools, only to swing them and realise the nail was never there.
I’ve been building AI products for about 5 years now. Before that, I was in lead generation, doing everything by hand: photos, landing pages, lead magnets, copywriting, the lot.
That manual work made me sharp. It taught me what good looked like and what actually moved the needle versus what just felt like it would.
When I first discovered no-code and automation, I didn’t try to automate marketing as a whole or "replace marketers". I used it to remove specific pains I was already feeling.
- Manually checking Facebook for fresh leads? Zapier could handle that.
- Need to split-test images and copy? There was a platform for that.
One real problem at a time.
Each solution was driven by friction I had personally felt. I knew what “done” looked like because I had done it manually or wrong a hundred times before.
Then I made a critical mistake: I confused leverage with value.
Automation felt infinitely valuable because it scaled almost for free. If something could be automated, I assumed that made it worth building at scale.
Elon Musk made a similar mistake at a much bigger scale when Tesla tried to automate too much of the Model 3 production line. He ended up walking it back with a simple admission: “Yes, excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.”
Looking back, what I was really doing was treating “can this be automated?” as the question, instead of “is this worth doing at all?”
The truth is simpler: value lives in the underlying workflow that solves a painful problem for someone. Automation is just the fastest way to run that workflow once you know it works.
I used to think in terms of:
automation × process = value
The real equation is closer to:
(validated problem × effective process) × automation = value
If the problem isn’t validated or the process doesn’t reliably create outcomes people care about, the left side is basically zero. You can throw infinite leverage at it and still end up with nothing. Leverage multiplies what’s already there. If there’s nothing underneath, you still get nothing back.
Before LLMs, I had an idea to automate buyer qualification for real estate agents: handle their SMS conversations, filter leads, and prospect lukewarm ones on autopilot.
What I should have done was the unscalable version: run the workflow manually, even if it meant hiring a couple of people or doing it myself. Prove that agents would hand over their buyers. Prove that the conversations actually improved qualification. Prove that the outcomes were good enough that they’d pay for it.
Instead, because I assumed the technology was the hard part, I built a platform led by the tech rather than the users. I spent months and tens of thousands on “technical proofs of concept” and shiny demos proving it could be done, but never proving it should be done.
Every feature was a guess. Every decision was based on conversations I imagined having. The roadmap was a list of untested assumptions with zero signal about which parts actually mattered, or what agents would pay for.
By the time there was something to show, months had passed in a feedback vacuum: no users, no data, no course corrections. Just me and a perfect machine wrapped around hypotheticals.
Meanwhile, someone else shipped something rough. They talked to real customers, manually solved painful problems, and then automated the bottlenecks. They iterated toward the truth while I was still polishing architecture. They got acquired by a billion‑dollar company. We ran out of money before helping a single user.
The lesson that keeps echoing in my head is:
0 × ∞ = 0
Leverage multiplies what already exists: a real problem, a working process, and sound judgment. Infinite leverage on zero traction still gives you zero.
So now, before building the machine, the question is:
if automation is a multiplier, what exactly am I multiplying?
All the advice about “doing things that don’t scale” suddenly makes more sense to me. It’s not anti‑automation. It’s just a reminder to start with reality instead of infrastructure. Start with the smallest complete version of what will prove your concept (AKA an "MVP") and once you know what works in the messy, high‑touch version, then—and only then—does it make sense to bring leverage to bear.
Action leads to information. Every plan is perfect in theory. You can’t learn to swim by watching YouTube videos; at some point you have to get in the water.
The only way to prove anything is to subject it to reality.