Build or buy: is it your company's way, or the world's way?

For thirty years, "build your own software" was the wrong answer for a small company: expensive, slow, and what remained was a system nobody knew how to maintain. That has changed enough for the question to come back — and it came back without an instruction manual.

The risk now is the opposite of the old one. With AI, building anything is cheap — and the temptation is to build everything, including what already exists, ready-made and better.

The question that settles most cases

Before any price comparison, there is one question that sorts cases into two piles:

is this the way your company works (build) or the way the world works (buy)?

Issuing an invoice is the world's way. Running payroll is the world's way. Antivirus, e-mail, accounting: the whole world does it the same, and whoever builds that is paying dearly to reinvent something already regulated and tested.

Now, how your company approves a request, the way it joins two pieces of information that live in different systems, the rule that only makes sense in your industry — that is yours. And that is exactly where no off-the-shelf product fits properly, because it was built for the average of a thousand companies.

Four paths, not two

The real choice is not between buying and building. There are four, and the order in which you ask matters:

Is it regulated, or a commodity the whole world uses the same way? Buy it ready-made — and never debate it.

Is there an existing service that solves nearly all of it, our way, at a reasonable price? Subscribe. But it goes in the inventory, with an owner and an annual review, or it becomes one more charge nobody can explain.

Is it integration between our own systems, a process done our way, or an edge no product does properly? Then yes, build it.

Does it demand deep, continuous expertise outside our focus? Heavy networking, telephony, cameras: a specialist supplier, closely governed.

Notice that "build" is the third branch, not the first. The tree exists to contain enthusiasm, not to authorise it.

The tie-breakers

When a case sits on the fence, three questions settle it:

How often is it used? You build what is used every day. What gets used once a quarter does not pay for the maintenance it will demand forever.

What does it cost over three years? Not today's price: accumulated subscription on one side, building plus maintenance on the other. Your own software does not end when it is finished.

Where does the sensitive data live? If the answer changes who answers for a breach, it weighs more than the other two.

Who directs and who executes

It is worth being precise about what AI does here, because the overblown promise has ruined plenty of projects: the human directs (architecture, judgement), the AI executes.

What got cheap was writing code. Deciding what should exist, what the system will not do, and what happens when it goes wrong still costs exactly what it always cost: the attention of somebody who understands the business.

And there is a technical choice that is really a business decision: use ordinary technology, the kind any competent professional can pick up. Your own software built with an exotic tool is not an asset — it is another supplier, this time with your name on it.

The step almost everybody skips

Here is the most expensive mistake, and it happens after delivery:

The delivered solution becomes "something that already exists" and is governed by the very system that created it. Nothing is left orphaned; the method feeds itself.

In other words: the day the system goes live, it stops being a project and becomes something the company owns — with an owner, with documentation, and with future problems arriving through the same queue as everything else.

Without that step, every solution built becomes an orphan: it works, nobody can explain it, and the day it breaks it will find the company with nobody responsible. That is exactly how the critical spreadsheets nobody dares to touch came to exist.

The test

Think of the next thing your company wants to solve with technology, and answer:

Is this the way your company works, or the way the world works?

If it is the world's way, buying is the cheap and correct decision. If it is your way, you already knew — and you are probably paying some subscription to do it badly.