- Most companies should buy and configure, not build from scratch.
- Custom development is worth it only for processes that are core to how you compete.
- The skill you need most is configuration and integration, not model training.
The question "should we build our own AI?" comes up in almost every boardroom. For most businesses the honest answer is no, at least not at the start. The better question is how much to configure.
Three levels of effort
| Level | What it means | When it fits |
|---|---|---|
| Buy | Use an AI feature inside software you already have, or an enterprise AI assistant | General drafting, summarising, searching documents |
| Configure | Build workflows on an automation or AI platform, connected to your systems | Repeatable processes such as enquiries, invoices, follow-ups |
| Build | Custom applications on top of leading AI models | Processes unique to your business that give you an edge |
Start with buy and configure
Off-the-shelf tools improve every few months, and you benefit without lifting a finger. Configured workflows on established platforms are quick to change when your process changes. Both keep cost and risk low while you learn.
When building makes sense
Build when the process is central to how you win business, when no product handles your specific rules or language mix, or when the volume is high enough that per-user licences become expensive. Even then, build on top of existing models rather than training your own.
The people question
Buying a tool is easy; making it work inside your processes is the hard part. Whichever level you choose, you need someone who can map the workflow, configure the tool, connect it to your systems and train the users. That is the skill set of an AI implementation lead or automation specialist, rather than a research scientist.
Want this working in your company?
Tell us how your team works today. Within two business days you get a recommended hiring model and a scoped proposal.
Talk to us