Custom software vs SaaS, how to choose for an Australian small business

The build vs buy decision in 2026 is not what it was in 2020, because AI has collapsed the cost of custom by sixty percent.

If you were an AU small business owner in 2019 and someone offered to build you a custom platform for forty thousand dollars, the right answer was usually "I will use Squarespace and Mailchimp instead." The maths punished custom builds. Off-the-shelf SaaS was almost always the winning choice.

The maths has changed. Senior developers with AI assistance now ship in twelve weeks what used to take eight months. The price for a real custom build has dropped from forty to fifty thousand dollars to fifteen to twenty-five thousand. That changes the build vs buy calculation for every AU SMB that runs a workflow on duct tape.

This post is the rewritten framework for 2026. The old framework. Why AI changes it. Three scenarios where SaaS still wins. Three scenarios where custom now wins. And how to decide in under an hour.

The classic build vs buy framework, abbreviated

The standard 2010s decision tree had three axes. Cost, control, and time to market.

SaaS won on cost. You paid forty-nine dollars a month, never a forty-thousand-dollar cheque. SaaS won on time to market. You signed up Friday and shipped Monday. Custom won on control. You owned the code, the data, the roadmap, and the ability to bend the tool to your specific workflow.

For the vast majority of AU SMBs, cost and time were the dominant variables. Control was a luxury you paid for only if your business was so distinctive that off-the-shelf could not represent it. Hence the default verdict, "use SaaS, save your money."

The framework was correct for its decade. Custom development was expensive enough that the cost overwhelmed the control benefit for almost everyone.

Why AI changes the maths

AI assistance changes one thing about custom development. It changes the cost.

In 2020, a senior AU developer building a custom B2B web app from scratch shipped roughly forty to sixty hours of usable feature work per week. That number was the binding constraint on the price. At one hundred sixty dollars an hour, a twelve-week build cost between seventy-five thousand and one hundred fifteen thousand dollars in direct labour. Plus project management. Plus design. The all-in number for a real platform was forty to eighty thousand dollars minimum.

In 2026, the same senior developer using Cursor or Claude Code ships closer to one hundred fifty to two hundred fifty hours of usable feature work per week, because AI assistance has compressed the boilerplate, the test writing, the data access layer, and the documentation. Senior judgement is still the binding constraint. The constraint has just become a smaller fraction of the cost.

The same twelve-week build now costs between twenty-five and forty-five thousand dollars all-in. ServoSimple, our own live AU petrol-station SaaS, cost roughly twenty-five thousand dollars in direct labour to build solo with AI assistance. You can read that case study at /blog/how-servosimple-was-built-case-study-shipping-saas-solo.

That is a sixty percent cost reduction. It does not mean every SMB should now go custom. It does mean the framework that automatically said "use SaaS" is now wrong fifty percent of the time.

Three scenarios where SaaS still wins

The SaaS default is still correct in three specific scenarios. If you are in one of these, do not custom-build.

Scenario one: the SaaS is a commodity and you do not want to think about it

Email. Payroll. Accounting. Calendaring. Document signing. Cloud storage. These categories are commodities. There is nothing you can build that meaningfully outperforms Gmail, Xero, Google Workspace, or DocuSign. Even if you could, the maintenance cost of the custom version eats the value gap immediately.

The honest rule: if every other AU business in your size band uses the same SaaS, you should too. Originality at the boring layer is a tax, not an asset.

Scenario two: the workflow is genuinely off-the-shelf and you are below the friction threshold

Some SaaS tools handle a workflow well enough that the friction of customising is real but not painful. CRM for a five-person sales team. Inventory for a single-warehouse retailer. Project management for a marketing agency under twenty staff.

If the workflow is roughly standard, and the friction you feel using SaaS is annoying but not blocking, stay on SaaS. The custom build cost is real, and at that scale the workflow does not yet justify the spend.

The honest test: can you describe the friction in a single sentence to your bookkeeper without them rolling their eyes? If yes, you are below the build threshold. Stay on SaaS.

Scenario three: you are pre-revenue or under three hundred thousand dollars annual revenue

The build vs buy decision is partly a financial question. At lower revenue, the opportunity cost of every dollar is higher, and the time to recover a build cost stretches past the runway. Stay on SaaS until the business has revenue stability and a workflow that has earned the spend.

The honest threshold: under three hundred thousand dollars in annual revenue, never build. Between three hundred thousand and one million, build only the single most painful workflow. Above one million, the calculation flips.

Three scenarios where custom now wins

These are the scenarios that flipped between 2020 and 2026. If you are in one of them, the build is now the right answer.

Scenario one: the SaaS tax is over fifteen percent of revenue

If you add up every SaaS subscription you pay each month, and the number is over fifteen percent of your monthly revenue, you have crossed the line where custom becomes cheaper. Every AU operator I have walked through this in the last year has been surprised by the total. The number adds up faster than you think when you have nineteen different tools on three different billing cycles.

A custom platform built once can replace six to twelve SaaS subscriptions for a flat monthly cost of hosting and maintenance. The break-even on a twenty-five thousand dollar build is usually under fifteen months if the SaaS tax is over fifteen percent.

Scenario two: your workflow is genuinely vertical-specific and no SaaS represents it well

ServoSimple exists because no SaaS represents the workflow of an AU petrol station. My Legal Connect exists because no SaaS represents the workflow of pairing legal contractors to law firms. Coiba Dental was built because no off-the-shelf dental site handled the international clinic's specific intake flow.

If your workflow is the workflow, and SaaS forces you to bend the business to fit the tool, custom is the answer. The cost of bending the business to fit Squarespace is higher than the build cost over three years. Almost always.

Scenario three: your team spends more than ten hours a week glueing tools together

This is the silent killer for AU SMBs. Operations staff burn ten to twenty hours a week copying data between spreadsheets, retyping orders from email to inventory, manually triggering email sequences, or reconciling exports from one tool against imports into another.

At forty dollars an hour of staff time, ten hours a week is twenty thousand dollars a year, every year. A custom build that absorbs the glue work has a payback period of under eighteen months and the saving compounds. The same SaaS-stack alternative does not get cheaper over time. It gets more expensive as you bolt on more tools.

How to decide in under an hour

Print this. Answer the questions. Score yourself.

Annual revenue. Under three hundred thousand, do not build. Between three hundred thousand and one million, only build the single most painful workflow. Over one million, custom is back in scope.

SaaS tax as percent of revenue. Under five percent, stay on SaaS. Five to fifteen percent, audit your stack but do not build. Over fifteen percent, build now.

Vertical specificity. If your workflow is genuinely your industry's workflow, and you cannot find a SaaS that represents it without compromise, build. Otherwise stay on SaaS.

Glue work hours per week. Under five hours per week, ignore. Five to fifteen hours, build a single internal tool to absorb the worst of it. Over fifteen hours, the build is the cheapest decision available.

If two or more of these point to build, the build is your answer. If two or more point to SaaS, stay where you are. The third axis is your tiebreaker.

How to start

If the answer above is build, the Voltari Digital Operator tier is built for this. Fixed price, fixed timeline, custom application functionality with AI integration included. We have shipped this exact build for ourselves multiple times, which is how we know what it actually costs.

Send a scope brief to /#contact. The first call is thirty minutes, no charge, and the output is a written assessment of whether build is the right call for your situation. Sometimes the answer is no, in which case we tell you which SaaS to use and where to invest the saving. The honest second opinion is worth a call either way.

The maths has changed. Most AU SMBs have not noticed yet. The ones who do will save twenty to forty thousand dollars in SaaS tax over the next three years and own the workflow at the end of it.