Once an AI use case proves its worth, a technology question follows: do you build it, buy it, or stitch existing pieces together? Get this wrong and you either reinvent the wheel or paint yourself into a corner. The right answer is rarely the same twice.

The three options

Each has its place:

• Build — worth it only when the capability is genuinely core and differentiating, and you can sustain it. Most things aren’t.
• Buy — fastest for common needs; you trade some control for speed and someone else’s maintenance burden.
• Orchestrate — combine existing models, platforms and tools into a workflow. Increasingly the pragmatic middle ground.

Choosing well

Be honest about what’s actually differentiating — most AI capability is now a commodity to buy or assemble, not build. Favour options that keep you flexible, because the technology is moving fast and today’s best model won’t be next year’s. And weigh the whole cost, including the people and effort to run it, not just the licence.

The pragmatic takeaway

Stay technology-agnostic and outcome-led. Pick the option that delivers the result with the least lock-in and the lowest total effort to maintain — and be ready to change your mind as the landscape shifts.