AI Economics
Build, buy or configure AI: choosing the right path
Why the answer is usually different for different use cases, and the commercial test we apply to every option.
By Whitesoft · · 6 min read
The question is rarely 'should we build or buy AI'. It is 'for this use case, what is the cheapest reliable way to get the outcome', and the answer changes from one use case to the next. We apply the same test to every option, and we frequently recommend the boring one.
Three paths, not two
- Buy: a vendor product that does the job out of the box, such as an AI feature in your CRM or contact-centre platform.
- Configure: a platform capability that needs your data, prompts and rules but no custom engineering.
- Build: custom development where the capability does not exist, or where it is central to how you compete.
The questions that decide it
| Question | Points towards |
|---|---|
| Is this capability how we differentiate, or a cost of doing business? | Differentiation favours build; cost of business favours buy or configure |
| Does an incumbent platform already offer it? | Buy or configure, unless quality is inadequate |
| How sensitive is the data, and where must it stay? | Constraints may rule out some vendors entirely |
| What is the three-year cost including licences, usage and people? | Compare honestly; vendor pricing changes |
| How hard is it to leave? | Prefer options where data and prompts are portable |
| Do we have people to run what we build? | If not, building creates a liability |
The mistake we see most
Organisations build what they should have configured, because building feels like progress and a vendor feature feels like settling. Six months later they own a system that a platform update made redundant. The reverse also happens: a strategically important capability is bought from a vendor who then owns the data, the improvement roadmap and the pricing.
Keep the option open
Whatever the path, keep your data, evaluation sets and prompts in a form you control. Model providers and platforms will change. The organisations that benefit are the ones that can move a workload when a better or cheaper option meets their quality bar.