Most build-versus-buy advice offers a generic weighing exercise: consider cost, consider time, consider risk, arrive at a balanced answer. That is fine as far as it goes, but it does not tell a business owner what they actually need to know, which is whether their specific situation shows the signals that reliably predict a custom build will outperform buying an off-the-shelf tool. Those signals are more concrete than a general framework suggests, and worth checking against directly rather than working through an abstract decision matrix.
Signal one: your workflow is genuinely unusual
Off-the-shelf tools are built for the median customer's version of a workflow. If your business's version of that workflow has a real structural difference, a compliance requirement specific to your state or industry, a data format no mainstream tool handles cleanly, a multi-step approval chain unique to how your business operates, an off-the-shelf tool will either not fit or will fit only after enough workarounds that you are effectively building your own process anyway, just with someone else's constraints layered on top.
Signal two: the volume justifies the setup cost
A custom build has a real upfront cost, typically somewhere between $1,500 and $10,000 depending on complexity for a well-scoped Claude Cowork skill. That cost only makes sense against a workflow with enough volume or frequency to pay it back inside a reasonable window, generally under a year for most Australian SMBs. A task run twice a month rarely justifies a custom build regardless of how well it fits; a task run fifty times a week almost always does.
Your workflow has a genuine structural quirk no mainstream tool handles cleanly.
The task runs often enough that setup cost pays back within about a year.
You already have, or can quickly get, someone capable of building and maintaining a simple Claude skill.
The off-the-shelf alternative's per-seat cost scales badly as your team grows, while a custom build's marginal cost per extra user stays low.
Signal three: per-seat pricing punishes growth
A SaaS tool charging per seat gets proportionally more expensive as a team grows, while a custom Claude workflow's marginal cost per additional user is close to the underlying API cost, far lower than a vendor's seat price. A ten-person team paying $40 a seat for a tool is at $400 a month; the same tool at fifty people is $2,000 a month for the same underlying capability. A custom build's cost curve looks nothing like that, which matters a great deal for a growing Australian business trying to model its AI spend two years out rather than just this quarter.
A worked example
A Melbourne wholesale distributor evaluated an off-the-shelf order-processing assistant priced at $85 a seat a month against building the equivalent workflow as a Claude Cowork skill. At the distributor's current team of eight, the off-the-shelf tool cost $680 a month, cheaper than the custom build's amortised setup cost in year one. Modelling forward two years against the distributor's realistic growth plan to eighteen staff, the off-the-shelf tool's cost climbed to $1,530 a month while the custom build's ongoing cost, mostly API usage that does not scale per seat, stayed close to $340 a month. The signals all pointed the same direction: real growth trajectory, a workflow with distributor-specific quirks the generic tool handled awkwardly, and volume that clearly justified the setup cost.
When these signals are not present, buy
If your workflow is genuinely standard, your volume is modest, and you do not have anyone available to build and maintain a custom skill, an off-the-shelf tool is very likely the right call and this checklist is not an argument against ever buying software. The point of these four signals is not to push every business toward building everything themselves. It is to replace a vague gut feeling about build-versus-buy with something closer to a checklist you can actually run against your own numbers before committing either way.
A fifth signal worth checking: maintenance appetite
A custom build is not a one-off cost. Workflows drift as your business changes, a new product line, a new compliance requirement, a new system you integrate with, and a custom skill needs someone willing to update it when that happens. An off-the-shelf tool's vendor absorbs that maintenance burden as part of the subscription price. A business with no appetite for owning that ongoing upkeep, even a modest amount, is often better served by a tool with a vendor behind it, regardless of how well the other four signals point toward building. Honest self-assessment on this point avoids the common failure mode of a custom build that works beautifully for six months and then quietly breaks when nobody has time to fix it.
Running through all five signals against a specific workflow takes about twenty minutes and produces a far more defensible answer than a general sense of which direction feels right. Do it before the next AI tool renewal, not after you have already signed another year's contract on a tool your business has clearly outgrown.
