You don’t need to understand transformers to make good AI tooling decisions. You need a framework. Here’s the one we hand out to clients.
The three options
Build: commission a custom solution. Highest cost, highest fit.
Buy: license an off-the-shelf SaaS tool. Lowest cost, lowest fit.
Wait: do neither. The category is moving fast enough that today’s decision may be obsolete in six months.
When to build
Build when the workflow is core to your differentiation, when off-the-shelf tools can’t reach the integrations you need, or when the data sensitivity rules out a third party. Custom builds work best when the workflow has been documented manually for at least 6 months — you know the edge cases.
When to buy
Buy when the workflow is undifferentiated (every business does it roughly the same way) and a mature category exists. Email marketing platforms, CRM, analytics — buy these. Don’t build a CRM.
When to wait
Wait when the category is volatile (model providers, agent frameworks), when you don’t yet have the data or process maturity to make the tool work, or when the cost of inaction is small.
The 4-question test
- Is the workflow core to our differentiation?
- Does a mature off-the-shelf category exist?
- Is our data and process mature enough?
- Is the cost of inaction high?
Most “Build vs Buy” decisions resolve cleanly from these four questions.