For months I’ve been wrestling with this question.
AI has changed the math. Describe the tool you need, feed it a good prompt, and two hours later you often have something that runs.
Sometimes it even looks polished.
That speed is intoxicating. Why hand over $9–$29 every month for a SaaS when you can build it yourself for “free”?
Then the second act begins.
The version on your laptop works perfectly. But you want to access it from your phone, or your tablet when traveling.
Deployment is next: hosting, domain name, SSL, environment variables, transactional emails, build pipeline.
Suddenly things don’t operate the same way.
What started as two hours quietly becomes several days of using AI to hunt and peck for solutions to issues as they arise.
If other users arrive, bugs appear instantly.
You patch one. Three more surface.
You are now the 24/7 maintainer of infrastructure you never planned to own.
You just wanted to build what you thought was a simple tool and look like a hero to friends who haven’t tried building yet.
And you discover it’s not just your code… every upstream library update or API change breaks something.
You spend evenings trying to get AI to see code updates and fix the issue you’re experiencing.
And you start thinking to yourself: the SaaS alternative you skipped? It already handles scaling, security patches, 99.9% uptime SLAs, rate limiting, GDPR/CCPA compliance, SOC 2 paperwork, and the dozens of edge cases you would have discovered the hard way.
Their engineers are on call at 3 a.m. You prefer to be sleeping at that time.
The true cost was never the initial two hours.
It is the cumulative hundreds of hours that follow. Those are hours you cannot get back.
Opportunity cost is the quiet killer here.
Every minute spent keeping a non-core tool alive is a minute stolen from the work that generates revenue.
Most months $10–$20 feels like an unnecessary line item. Until you do the real math and realize it buys back focused time you can actually sell, use to grow your net worth, or simply just relax.
I still prototype aggressively with AI, especially with V0 and Bolt. That part remains magical and I recommend it.
But the moment the prototype must become a long-lived, reliable piece of my business, I start to reach for my credit card almost every time.
The decision rule I now use is brutally simple: If the tool saves me even one high-quality focused hour per week, the subscription more than pays for itself.
If it doesn’t, I probably didn’t need the tool in the first place.
Build when it creates (or could create) an unfair advantage for you.
Pay when someone else has already solved the plumbing far better than you ever will.
That’s where I’ve landed after too many late-night debugging sessions.
###