Build, Buy, or Build

For decades, software had two doors.

Most businesses chose Door #2 because building software was expensive, slow, and risky. The Standish Group put project success at 31%.

So you bought the CRM, the accounting, the email platform, the scheduler, helpdesk, analytics, payments—all of it.

And now? Too many subscriptions, using a fraction of the features, and barely auditing what you spend. My business has 14 paid subscriptions that I know about.

So I should build instead?

Not exactly. SaaS fatigue is pushing people toward the wrong conclusion. There aren’t two doors anymore.

There are three.

Let’s move through this quickly, because Buy is still the right answer plenty often.

That logic still holds.

What changed is the AI layer inside purchased software. HubSpot, QuickBooks, Shopify, Google Workspace—they’re all shipping capabilities that would have felt enterprise-only a few years ago. You can win with SaaS by learning the AI features.

The pattern worth remembering is simple: buy the platform, build the last mile.

That last mile—the customization layer on top of purchased systems—is increasingly where the advantage lives.

This is real software: architected, tested, maintained, relied on for years.

AI has changed the economics of the first version:
Version 1.0 got cheaper.
Version 1.47 didn’t.

Maintenance → still shows up every year.
AI-generated code → accumulates technical debt 4x faster by year two.

Forrester predicts 75% of tech leaders will face moderate to severe technical debt from unstructured AI code generation — not in the future, but right now.

"So AI makes it cheaper to build but not cheaper to keep?"

Bingo!

Traditional Build got faster. It did not get casual.

This is fast prototyping, and you don’t even have to be technical. AI wanted me to shorten that list, but building with AI is too much fun!

The bottleneck was never just code.

It was translation — getting what you know about your business into software. That used to require engineers. But now, the person closest to the problem can build the solution.

It’s kinda obvious.

A distributor with 15 years of quota assignments > a developer who's never sold anything trying to guess how the logic works.

The advantage lives with the person who understands the workflow.

Anish Acharya at a16z put it this way: "Software creation used to be constrained by ROI. Now it's constrained only by imagination — and that's a much more interesting limit."

Too small, too temporary, or too specific? Not anymore.

You can build a rough version in hours, learn what matters, then:

🗣️ So Go Build with AI!!

Action produces information.

Seriously, what are you waiting for? Go build something! Come back if you must.

The old question was: Which SaaS should we buy?

The real question is:

Which requires discernment (my favorite word these days).

The role is emerging but unnamed.
Someone who sits at the intersection of AI fluency, business operations, and honest assessment.
Someone who can look at your workflow and say: buy this, build that the old way, build this with AI, and stop paying for that…….

If your business is at the point where you need someone to help you sort out which door is which — that's what Wink Intelligence does.

The risk is building outside your domain knowledge.

That distributor building a quota tracker? ✅
Vibe-coding a payment gateway? ❌

The disasters in this space follow a pattern: people use disposable tools to do production-grade work. They end up with security failures or broken systems.

(if you need an AI policy at your SMB, try my free AI Policy Generator)

Those disasters aren't failures of disposable software. They're Traditional Build attempts, misdiagnosed, because “anyone can code now.”

A few tests worth running.

Audit first. Do you even know what you're paying for? Or the capabilities?

Core or commodity? Does this make or save money in a way your competitors can't replicate? ✅ — you're in Build territory. But which kind?

The four-question test. Narrow scope? Short lifespan? Single owner? Low stakes if it breaks? ✅✅✅✅ → Build with AI

The sick test. If you got sick for a week, could someone else figure this out?

Every business has knowledge, workflows, and operating logic that live inside its walls and nowhere else.

The thing that connects your CRM to → your comp plan → your Monday meetings → your handoffs between departments… That's yours. No vendor built it because no vendor could. It lives inside your walls.

Your business runs on specifics, and you can teach AI those specifics.

Every business owner right now has the same quest - making AI truly useful.
The business application phase is here (for a couple years, I think) - so build that last mile, build disposable, or build to learn!

If this helped you think about your own build-vs-buy decisions, share it with someone else who's wrestling with the same question.

And if you want help with strategy, training, or custom agents, reach out:

I translate big concepts in 900 1100 words or less. I should probably do video or podcast, but I like to write.

Thursdays: We’ll cover LLM and AI concepts. (Don’t worry. I put the cookies on the lower shelf, where anyone can reach ‘em.) I teach advanced concepts in simple, bite-sized chunks that anyone can understand.