For years the decision was simple. If software existed for your problem, you bought it. If it didn't, you paid a developer, or lived without. AI has quietly rewritten that maths, and plenty of businesses are still paying prices from the old one.
Here is the version of build-or-buy we walk owners through now.
What changed
Building used to mean hiring. An agency owner at one of our Melbourne workshops had two quotes in hand for a client project: $50,000 from one developer, $2,000 up front plus $1,000 a month from another. Then she built the first working version herself, in the room, that morning.
Stories like hers keep arriving. A deck builder came through and built a 3D calculator that turns a customer's measurements into an estimate. Another attendee, three hours in, had a booking confirmation system talking to his own CRM. None of them came in as programmers. The cost of building the small, specific tool your business needs has fallen a long way, and that changes which side of the ledger many problems belong on.
Still buy these
Buying remains right where the problem is universal and the stakes of getting it wrong are high. Accounting. Payments. Payroll. Your core CRM. These are solved problems, maintained by companies whose whole job is keeping them compliant and running. Rebuilding them yourself is a hobby, not a strategy.
A rough test: if thousands of other businesses have exactly your problem, buy. Your quoting rules and your onboarding steps are a different matter. Off-the-shelf products were not written with those in mind. That is usually build territory.
Build the glue
The most valuable builds are rarely whole products. They are the glue between the tools you already own: the enquiry that becomes a CRM entry without anyone typing it, or the report that assembles itself every Friday.
Vendors struggle to sell you this glue, because it is specific to how your business runs. It is also what owners at our workshops tend to build first, because it eats the most admin time. One owner described his business juggling four or five separate platforms with staff re-keying data between them. Pointing AI at that re-keying was his first project. Glue sits at the small end of the effort scale, and workshop attendees regularly get a first version running inside the day. Full systems are a bigger job. One builder we know estimated a complex client build at around 40 hours.
The tool-stack trap
The buying instinct has a failure mode worth naming: the stack. Businesses under pressure buy tools the way anxious people buy gym memberships, and for the same reason. Each purchase feels like progress. Somewhere down the track there is a subscription list nobody audits and a team using a fraction of each tool.
Before any purchase, one question: which named job does this remove? If the answer is a feature list rather than a job, it is not a purchase. It is a postponement.
Ownership is the tiebreaker
When a problem genuinely sits on the line between build and buy, we weight one factor above the rest: ownership. A tool you build runs where it suits you, on your own machine or a server you control, and it changes when your business changes. A tool you rent changes when the vendor changes it, prices included.
Being able to make your own tools is a recent development for most owners, and it is the quiet heart of the build-or-buy shift.
A fair way to decide
Where the problem is universal and the stakes are high, buy. Where it is specific to you or glue-shaped, build. And when you genuinely cannot tell, a day spent trying to build it will often settle the question. Even a failed attempt tends to teach you what you actually need.
The maths on building has moved. Before the next software renewal lands, it is worth an hour deciding which side of the ledger each subscription now sits on.
