Buy vs Build vs Fractional: A Decision Framework for Enterprise AI

The three options, stated properly

Buymeans procuring a vendor product that solves a specific AI problem. You configure it to your domain. You don't own the model, the infrastructure, or the roadmap.

Build means hiring engineers (or an agency) to create a bespoke system. You own the code, the data pipelines, and the ability to evolve it.

Fractionalmeans bringing in an interim senior leader who defines the roadmap, sets up the function, hires the permanent team, and hands over. You don't own the person. You own what they leave behind.

Buy when the problem is solved and commoditised

• 1.Multiple credible vendors address the use case.
• 2.Your domain is not a meaningful differentiator. If competitors buy the same tool, it's infrastructure, not moat.
• 3.Speed to value matters more than ownership.

Where buy breaks: vendor capabilities plateau at the 80% that serves most customers. The last 20% — the part your business cares about most — often requires customization the vendor won't prioritise for you.

Build when the problem is your business

• 1.The workflow is a strategic core process.
• 2.Your data is the moat. The value compounds with your accumulated data.
• 3.You need integration depth that vendors can't provide.

Where build breaks: you end up owning infrastructure you don't have the team to maintain. “Build” without a sustained ops capability is how enterprises end up with five ML models in production that nobody has looked at in eighteen months.

Fractional when the gap is leadership, not labour

• 1.You have engineers, but no senior AI leadership.
• 2.A permanent hire at the right seniority would take 6+ months.
• 3.The work is one-time or cyclical. Building a team, defining an operating model, shipping the first initiatives.

Where fractional breaks: if the intent is to avoid hiring permanently while getting permanent-level output. Fractional works when the exit condition is clear — a permanent team in place, a roadmap handed over.

The decision matrix

Where the combinations live

Fractional leader + build team. An interim Head of AI defines the roadmap, hires permanent leaders, sets up the first two projects, and hands over at month 9-18. The organisation ends with a durable AI function and two shipped systems.

Buy + build. A vendor tool handles the commodity 80%, a custom layer handles the differentiating 20%. Neither alone would work.

Fractional + buy. The fractional lead evaluates vendors, avoids the expensive procurement mistake, and lands the right one with integration support.

A final honest note

I run USQRD. We do fractional leadership, and we do scoped build engagements. We don't do both at the same client at full depth, because the incentives conflict — the same people setting strategy shouldn't be the only ones bidding on execution.

If you're at the stage where you need a vendor, the honest answer is to talk to vendors, not to us. The framework is more useful than the vendor pitch. Use it.