Build vs buy AI: a decision framework
Buy when a product covers about 80% of the need and the remaining 20% is not your differentiator. Build when the process or the data is what makes you different, when integration depth decides the outcome, or when residency and audit rules rule the product out. Integrate when both are true: buy the platform, build the layer that is yours.
Buy
A product that covers the need today, with a subscription, a roadmap someone else owns and limits you inherit.
Build
Software shaped around your process and data, owned by you, with the running cost and the maintenance that come with ownership.
Verdict by criterion
| Criterion | Buy | Build |
|---|---|---|
| Time to first value | Days to weeks | 6–16 weeks |
| Fit to your process | You adapt to the product | The software follows the process |
| Ownership of data and logic | Vendor's terms | Yours, including prompts and weights |
| Cost at low volume | Lower | Higher |
| Cost at high volume | Per-seat or per-resolution pricing compounds | Fixed build, controllable running cost |
| Residency and audit control | Whatever the vendor offers | Private deployment possible |
| Maintenance burden | Vendor's | Yours or your partner's |
| Differentiation | Same capability as competitors | Capability competitors cannot buy |
Five questions that settle it
- Does an existing product cover 80% of the need without changing how you work?
- Is the remaining 20% the thing customers choose you for?
- How deep is the integration with systems only you have?
- Do residency, audit or sector rules exclude the product?
- What is the cost of each at your volume in three years, not this quarter?
The third option people forget
Integrate. Buy the platform and build the layer that is yours: a copilot on top of a bought CRM, an extraction pipeline feeding a bought ERP, a retrieval layer over a bought help centre. It is the most common right answer, and the one vendors on both sides argue against.
What it costs to choose wrong
Building what you could have bought spends six figures on a commodity and a maintenance burden that never ends. Buying what you should have built caps your product at the vendor's roadmap and prices your differentiator per seat. A two-to-four-week strategy engagement answers it with your numbers; see AI Product Strategy.
Frequently asked questions
Do you ever tell clients to buy?
▾
Yes. When a product covers the need, implementing it is cheaper and faster than building, and we say so in the discovery readout.
Can we start bought and build later?
▾
Often the best path: prove the workflow on a product, then build the part that became your differentiator, with the data you now have.
What about building in-house versus with an agency?
▾
A separate question. In-house wins on continuity, an agency on speed and having solved it before; most companies blend the two.