Ad hoc software: why companies are moving past off-the-shelf

For two decades, "buy off-the-shelf" was the safe default, and it was the right call. Building custom meant long timelines, large budgets and real risk of ending up with nothing. Buying meant being operational next month.

Most companies made that decision once, years ago, and never revisited it. Which would be fine — except the arithmetic underneath it has changed.

What actually changed

AI accelerates every stage of building software: reading an existing codebase, designing the solution, writing the code, generating the tests, wiring up the integrations. What used to take a quarter now takes days.

That doesn't make custom software free. But it moves the threshold. Systems that in 2022 needed a six-month project and a team of five now need six weeks and a small one — which puts them within reach of companies that could never have justified them before.

The interesting part isn't that building got cheaper. It's that the minimum viable size of a project got smaller. A process worth automating used to have to be big enough to justify six months of effort. Now a process that costs one person ten hours a week is enough.

What didn't change

Understanding the business.

Someone still has to sit next to the person doing the work and discover the step that appears in no manual, without which none of the numbers reconcile. Someone still has to decide whether a returned order subtracts from the day it was placed or the day it came back. Someone still has to stay for the three weeks when the new system feels harder than the old one.

None of that speeds up with models. And it's the part that determines whether the project is useful or merely finished.

Which means the balance of a software project has shifted: the code is no longer the expensive part. Understanding and adoption are. Any vendor still pricing a project purely by lines of screens hasn't noticed.

Off-the-shelf versus ad hoc

Generic SaaS asks your company to adapt to the tool. Ad hoc flips it: the software adapts to you.

The practical differences show up in five places:

• Features. You stop paying for what you never open — and more importantly, your team stops navigating around it.
• Roadmap. What you need gets built because you need it, not because a thousand other customers voted for it.
• Process. The system models how you actually work, including the step that makes you better than your competitor.
• Integration. Native connections to what you already run, instead of a connector that covers 80% and abandons you on the rest.
• Ownership. It's an asset on your books with the code in your hands, not access that can be repriced or discontinued.

Where it pays off

Custom earns its keep in three situations:

1. Processes core to your competitive edge. If how you do something is why customers choose you, a generic product flattens it to the market average.
2. Repetitive manual work that leaks hours every week. The easiest return to calculate and the fastest to see.
3. Integrations between systems that "don't integrate." Usually the highest-value work, because the data already exists — it just can't move.

Not all or nothing

Anyone framing this as a binary is selling something.

Keep SaaS for the commodities: email, accounting, payroll, e-signature. Build ad hoc where your differentiation lives, and connect the two so data moves without being typed twice.

We work that way ourselves. Our CRM is custom because no product modelled our funnel. Our accounting runs on something we bought, like everyone else's, because invoicing isn't our competitive advantage.

The honest caveat

Ad hoc is the wrong answer more often than a studio like ours would like to admit.

If the process is standard, a product does it better and cheaper. If the business is still finding its shape, building now freezes a snapshot that will age badly. And if the real problem is that nobody has decided who owns the process, software won't fix it — it'll just make it visible.

The point of building isn't fewer people. It's the same team focused on what matters, with software that's genuinely theirs.

If you want to know which side of that line your case falls on, that's exactly what a diagnosis answers — including when the answer is "don't build."