There are two ways a small business gets this wrong, and they look like opposites.

The first: the thing that makes the business different — the way it quotes, the way it schedules, the sequence its jobs move through — gets handed to a subscription built for the average version of that business. It fits badly. Everyone develops workarounds. Within a year the conclusion is "software doesn't really work for us", when what actually happened is that the company's one distinctive habit got filed down to fit somebody else's default settings.

The second: the business builds its own accounting, or its own leave tracker, and spends three years maintaining a worse version of something that costs a few dollars a seat and updates itself when the rules change.

Both failures come from asking the wrong question. "Build or buy" gets argued as a price comparison, and price is the least useful thing on the table. The question that decides it is positional: where does this job sit relative to what makes you different?

Where custom software for Singapore SMEs actually belongs

Almost nothing is purely build or purely buy. There are three places a job can live.

The job is a commodity. It runs the same way in your business as in three thousand others. Accounting, GST returns, payroll and CPF, email, storage, e-signatures. There is nothing distinctive about how you do these, and you do not want there to be. Buy them. Buy them cheerfully — you are also buying somebody else's obligation to track the rule changes, which is worth more than the licence. When InvoiceNow moved e-invoicing onto PEPPOL, businesses on maintained accounting software got the format handling as an update. That is the whole value proposition of buying, and it is a good one.

The job is nearly a commodity, with one twist. A standard CRM, except your quotes follow a pricing rule the tool has no field for. Off-the-shelf stock control, except your reorder logic depends on something the vendor never imagined. This is where most Singapore SMEs actually live, and the first move is to push the tool as far as it goes — custom fields, the automation layer, the integration. Extreme configuration is underrated and often ends the discussion.

The honest limit: you keep the vendor's maintenance, but the box is still theirs. Your configuration is a tenant in someone else's building, and every major upgrade is a negotiation you do not control. Survivable when the twist is a convenience. Not survivable when the twist is the business.

The job is the edge. This is the one people misclassify most often, usually downwards. If a customer picks you over the firm down the road because of how you do this specific thing — your inspection workflow, your scheduling method, the reason your turnaround beats theirs — then that job is not a back-office function. It is the product, whatever your UEN says you sell. Handing it to generic software means slowly becoming generic, and the erosion is invisible while it happens because every individual workaround looks reasonable.

That is where custom software for Singapore SMEs earns its keep, and nowhere else. Not because bespoke is better. Because that specific job cannot be bought without being changed.

The cost question, answered honestly

Every discussion of software development cost for a Singapore SME reaches for a comparison that does not hold: the subscription's monthly figure against the build's one-off figure. The shapes are different, and the shape matters more than the size.

Buying is small, recurring, and forever. It does not end, it rises when the vendor decides it rises, and it carries a second cost nobody puts in the spreadsheet: the work of bending your process to the tool, paid every week by the people doing the bending. Usually worth it — for a commodity job, worth it without reservation. But it is rent, and rent does not stop.

Building is one setup you own, then the discipline of keeping it correct. The setup ends. The discipline does not — software that nobody maintains rots quietly, and the people who say "build it once and you're done" have not run anything for five years. What you get in return is that the workflow bends to you, and the thing you are good at stays the thing you are good at.

Here is what genuinely changed, stated carefully because this is the part most often oversold. Custom software used to be a big-company purchase — not because small companies did not need it, but because the setup cost was priced for balance sheets that could absorb it. A five-person business could not justify a build, so it bought the generic tool and lived with the bend.

AI-assisted development collapsed that setup cost. We know because we did it to ourselves: in October 2024 we had seven offshore developers, and by January 2026 we had none, with everything driven in-house. I ship bug fixes and change requests five to seven times faster than I did under the old offshore arrangement — my own figure from my own delivery calendar, not a study. The full sequence is here, including the parts that went badly.

For an SME owner that means something narrow: the threshold moved. Builds that were obviously irrational at your size three years ago are now worth a serious afternoon of thought. It does not mean building is cheap, and it does not mean you should build what a subscription already does well. The maths of who does what changed; the positioning question did not.

Four questions, one evening

Run each candidate job through these. Answer honestly rather than optimistically — the optimistic answers are how businesses end up maintaining their own leave tracker.

1. Is this job written down? Not "we all know how it works" — written down, specifically enough that a new hire could follow it. The honest read: if it is not written down, stop here. You do not have a build-or-buy decision yet; you have a documentation problem, and buying software will encode the confusion at speed while building it will encode the confusion permanently. This is the same first gate as the readiness checks, for the same reason.

2. Do three products already cover most of it? Not one vendor's claim — three you could genuinely shortlist this week. The honest read: if three exist, the job is a commodity and the market has already paid to solve it. Buy. If you cannot find three, that is real evidence the job is less standard than you assumed.

3. Would bending your workflow to the tool cost more than building the tool to your workflow? The honest read: it depends on where the bend lands. A bend in how you file expenses is free. A bend in the thing customers choose you for is not a workaround, it is erosion — and its cost shows up as lost distinctiveness two years later, not as a line item now.

4. Who maintains it after launch? Name them. Not a department — a person, for the version of this software that exists in eighteen months. The honest read: if you cannot name anyone and you are not paying someone whose job it is, build nothing. This question retires more custom projects than price ever has, and it should. A build with no maintenance path is a subscription you pay in outages instead of dollars, and it is the gap most businesses discover late.

A job that passes all four — written down, no ready-made market, a bend that costs you where it hurts, a named owner — is a build. A job that fails any one of them usually is not, and the failure tells you what to fix first.

What we do

NEXT3LABS is an AI-powered engineering studio in Singapore. We build and run bespoke software for small and mid-sized businesses — designed to your workflow, maintained after launch, with AI-assisted development doing the volume work so the setup stays within reach of a company your size. We replaced a seven-person offshore development team with AI-assisted development to make that possible.

The part worth saying plainly: we will tell you when the answer is buy. Most jobs in most businesses should be bought, and a studio that recommends a build for every enquiry is selling hours rather than judgement.

So take the job you would most like to fix and ask it the four questions. If it is written down, three products already cover it, and nobody would notice the difference — buy the subscription this afternoon. If it is the job your customers actually choose you for, and you have been quietly working around a tool for two years, that is the conversation worth having.

If you want a second opinion on which side of the line your job falls, the contact page is a real person, and the conversation costs nothing.

---

NEXT3LABS builds and runs high-stakes custom software from Singapore. We write up what actually happened on our own projects — including the parts that went wrong.

Working on something where being wrong is expensive? Message us on WhatsApp — no pitch, happy to compare notes.

More notes · Other ways to reach us