Framework · How to think about it
Build, buy, or point AI at it
Every operational problem now has three plausible answers, and each comes with its own sales pitch. This framework sorts them by the shape of your problem so the decision is yours, not the vendor’s.
The core idea
Ten years ago the question was build or buy. Now there’s a third door “can’t AI just do this?” and the three get argued as competing philosophies. They’re not. They’re tools for differently-shaped problems.
Buying wins when your process is standard. Building wins when your process is the advantage. AI wins when the bottleneck is repetitive cognition reading, sorting, drafting at volume. Most real situations call for a combination, anchored by one honest question: what, specifically, is the expensive part?
The framework
Three answers, three problem shapes
Answer 1
Buy
A configured off-the-shelf tool the right default when your process is close to standard for your industry.
Choose this when
- ✓A configured tool covers ~80% of the job today.
- ✓Your process is fairly standard for your industry.
- ✓Speed to value matters more than perfect fit.
Answer 2
Build
Custom software shaped around your operation for when the workflow itself is how you win.
Choose this when
- ✓The workflow no vendor models well is your edge.
- ✓You’re paying people to bridge gaps between tools.
- ✓You need to own and evolve the system for years.
Answer 3
Apply AI
AI aimed at a specific workflow step powerful for repetitive cognition, oversold for everything else.
Choose this when
- ✓The bottleneck is reading, sorting, or drafting at volume.
- ✓Rules can’t capture the variation, but a human can review.
- ✓The data the AI needs actually exists and is accessible.
FIG. 01 The same problem, sorted by shape: standard process, differentiating workflow, or repetitive cognition.
How to decide
Name the expensive part first
Before any vendor call, write down the single most expensive step in the workflow in hours, errors, or missed revenue. If a standard tool handles it, buy. If it’s expensive precisely because it’s yours, build around it. If it’s expensive because skilled people do repetitive reading and re-typing, that’s the AI-shaped piece. And if you can’t name the expensive part yet that’s the real first problem, and it costs a conversation, not six figures.
Common questions
Build vs buy vs AI FAQ
Can the answer be more than one?
+
Usually it is. The most common recommendation we make is a hybrid: buy the standard core, build the thin custom layer that carries your advantage, and apply AI to the repetitive cognition inside it.
Isn’t AI just part of building?
+
Technically yes, but it deserves its own column because its decision profile is different: probabilistic output, review workflows, and data readiness questions that traditional builds don’t have.
What if we choose wrong?
+
Phasing is the insurance. Start with the smallest piece that proves value a configured tool, one custom module, one AI-assisted step and let evidence, not architecture debates, decide the next move.
Three doors, one honest answer.
Bring us the problem. We’ll tell you which door it actually points to even when the answer isn’t us.
