For about twenty years the answer was buy. Not as a considered position. As arithmetic. Custom software cost hundreds of thousands of dollars, took a year and had a well-earned reputation for arriving late and not quite fitting. Against that, almost any package that got you to eighty per cent was the rational choice.
The arithmetic moved. A competent team with AI tooling now builds in weeks what used to take quarters, which means the threshold at which building is worth considering has dropped a long way down. Plenty of businesses that were correctly told "you cannot justify custom software" in 2019 would get a different answer today.
That is the interesting development. It is also where people start making a new mistake, because a cheaper option is not automatically the right one.
When buying still wins
The problem is genuinely common. Payroll. Accounting. Email. Thousands of businesses need the same thing, the compliance burden is heavy and moves. A vendor amortises that across all of them. Building your own is not brave, it is expensive in a way that keeps being expensive.
The rules change without asking you. Anything downstream of tax, awards or regulatory reporting. Somebody else tracking those changes and shipping the update is most of what you are paying for.
It is not how you compete. No client has ever chosen a supplier because of their leave approval workflow. Buy the adequate thing, spend the attention elsewhere.
When building wins
The process is how you compete. Every business has one or two things it does differently and that difference is usually why clients stay. Forcing it into a package flattens exactly the thing you are good at. This is the strongest argument for building and the most frequently ignored.
You are paying for a lot of software to use a little of it. A licence per seat for a platform where your team touches four screens. Worth actually working out the annual figure before assuming it is cheaper.
You are already paying for the gap. The package does eighty per cent. The other twenty is a spreadsheet, a shared mailbox and someone re-keying between two systems for six hours a week. That is not a gap, it is an ongoing cost that never appeared in the business case.
Integration is the whole job. When what you need is mostly getting four systems you already own to agree with each other, a fifth product rarely helps.
The third option people skip
Neither.
It is the outcome we reach more often than either of the others and it is always the cheapest.
The feature exists and is switched off. Genuinely routine. Businesses run for years around a limitation their current platform removed in an update nobody read.
The process should not exist. A report assembled monthly that three people receive and none read. A reconciliation between two systems that only disagree because of a rule nobody has revisited since 2018. Automating either is building a faster version of waste.
A decision would dissolve it. Most of the complexity we are asked to build around is accumulated exception handling. Somebody senior choosing one way of doing something removes more cost than any amount of software. It also takes an afternoon.
How to actually decide
Skip the feature comparison. It is theatre. Every package wins it on its own website.
Map the process first. The real one, with the workarounds and the exceptions and the actual hours. Then three questions:
- Is this how we compete, or is it plumbing? Plumbing gets bought.
- Does the shape of it match what is on the market, or does it only match us? Only us leans towards building.
- What is the honest annual cost of doing nothing? Sometimes that number is small, which is the most useful thing you can learn.
An hour of that beats a fortnight of demos.
The bit that has actually changed
Not that building is now the right answer. That it is now a real answer, available to businesses for whom it was previously theoretical.
Which puts more weight on deciding well, not less. Cheap software you did not need is still money spent and it comes with something the expensive kind never did: enough people building things quickly that the question of whether the thing is any good has become the hard part.