Build, buy, or partner: an AI decision framework for healthcare groups

The pitches arrive constantly. Clinical decision support, scheduling automation, revenue cycle tools, procurement intelligence, patient communication platforms. Each one comes with a strong case for why it matters and a company ready to explain how the capability could be built, bought, or partnered into the business. The natural next step is to pick a path, and that is exactly where most operators slip.

Build, buy, or partner is the second step of the decision, not the first. The first step is classifying what kind of capability it is for the business. Most operators jump straight to the path, which is how groups end up building things they should have partnered for and buying things that needed deeper customization than off-the-shelf software ever delivers.

Build, buy, or partner

The three paths are older than AI, and their shape is straightforward. Build means owning the team, the technology, and the outcome. It offers the most control, takes the longest, and carries the highest cost, but the capability becomes genuinely the company's. Buy means licensing a finished product from a company, which is the fastest path and the cheapest upfront, with the tradeoff that the buyer lives inside whatever the product happens to offer. Partner means working with an outside specialist in an ongoing way: the specialist brings depth the buyer cannot replicate, and the buyer brings scale, data, and operational reality the specialist needs. It is more tailored than buying and less burdensome than building, but it comes with its own tradeoffs: dependence on the partner's roadmap, lock-in around shared data, and the work of managing a relationship rather than a contract.

Core versus context

Geoffrey Moore, a longtime advisor to Silicon Valley executives and the author of Living on the Fault Line, introduced a useful distinction. Work in a business is either core or context, and the two are not the same thing as important and unimportant.

Core work is what makes customers choose a group over its competitors. For a multi-location healthcare group, core work usually lives in clinical outcomes, the provider experience, and the customer relationship: all the things a returning customer would actually point to. Context work is everything else that has to get done well but does not differentiate. Accounting is context. Payroll is context. Building maintenance is context. None of it is unimportant.

Between them sits a middle category: work that matters strategically but takes specialized expertise, the quadrant Moore says to hand to a partner; for shorthand, this piece calls it core-adjacent. Cybersecurity is the classic example for almost every industry. Executive recruiting is another. Healthcare procurement is a third.

The rule

How the capability is classified determines the right path. Build what defines the group to its customers. Buy what is essential but does not set the group apart. Partner where the work matters strategically but takes specialized expertise. Get the classification wrong, and the result is wasted budget, distracted teams, and tools that never quite fit.

Where the rule breaks

Misclassification usually happens in three common ways. The first is treating every important capability as core: payroll has to work, but it is not what makes the group different. The second is assuming high stakes mean the work belongs in-house; clinical AI is the case in point: high-stakes, clinically sensitive, and almost every leading group partners for it. The third is letting urgency win over classification: speed pressure pushes leaders toward whatever feels fastest, which is almost always buying, and the mismatch shows up two years later.

Cost, urgency, and what the team can already do are real factors, but they come after the classification. They change the timing, not the answer.