When Building Got Cheap
Two years ago, automating a single ETL flow took me weeks.
I’d copy-paste code from ChatGPT into VS Code, wrestle with APIs, break things, fix them, repeat. Shipping anything end-to-end felt heavy.
This week, I built a full internal tool. UI, logic, data model, AI-assisted extraction — in five days.
And today, I was pitched a $400 AI “agent” to do something I already have on my roadmap.
I laughed. Not because it was bad. But because it revealed something uncomfortable.
The real shift isn’t better AI tools. It’s the collapse of the cost of problem-solving.
Here’s the lens I now use for every SaaS demo I sit through:
Can I build this myself? And if so — how fast?
That question is lethal to most AI SaaS.
Because a lot of software businesses were never selling outcomes. They were selling time savings when building was hard.
That world is gone.
The old build vs. buy SaaS equation just doesn’t hold up anymore.
When even a $25/user/month tool no longer makes sense past a certain headcount, the entire premise of “buy instead of build” is flipped.
It’s not just that building got cheaper. It’s that the math of “buy” stopped making sense once you realize you can spin up your own fit-like-a-glove solution in days.
Today, a domain expert with clarity, coffee, and AI can: design the system, generate the code, build the UI, own the data model, and ship end-to-end — without a team.
I’m not special. I’m not doing anything nobody else can do. That’s the point.
AI didn’t just lower costs. It desecrated the barriers to problem-solving.
Which is why so many AI tools are quietly priced like consulting. They don’t have moats. They have demos.
The real moat now is: deterministic systems, owned data models, boring correctness, AI at the edges — not the core.
Infrastructure beats vibes. Every time.
If you’re wondering why SaaS demos don’t land the way they used to, this is why.
The question isn’t “Is this impressive?” It’s “Is this better than me + AI, in a week?”
Most aren’t.
And that changes everything.