Should your college build its own AI platform?

Every college with a competent computer science department eventually asks the same question: why buy this when our own faculty could build it? It is a fair question, and it deserves a more honest answer than most vendors give.

Start with what building actually means. A working campus assistant is not a wrapper around a model API. It is a retrieval pipeline over your circulars and syllabi, an ingestion path that survives scanned and handwritten records, a permission layer that stops one department reading another's confidential material, an audit trail that can be produced during an inspection, and an evaluation harness that tells you when answers have quietly degraded. Each of those is unremarkable on its own. Together they are eighteen months of sustained work.

The cost that institutions consistently underestimate is not the build. It is the second year. Models are deprecated, embeddings need regenerating, a regulation changes and the retrieval logic has to change with it, and the two capable people who built it have both moved to industry salaries you cannot match. Software that nobody owns does not sit still; it rots.

There is also an opportunity cost that rarely appears in the proposal. The faculty capable of building this are the same faculty capable of publishing, supervising projects and teaching well. Redirecting them into platform maintenance is a real institutional trade, and it is usually a bad one.

Buying has its own failure mode, and it is worth naming plainly. A per-seat consumer subscription puts your institutional documents on someone else's infrastructure, prices you higher every year you succeed, and leaves you with nothing if you stop paying. That is not a purchase; it is a tenancy.

The position worth holding is narrower than either extreme: buy the platform, own the asset. The institution should end up holding its own index, its own documents and its own embeddings, running inside its own perimeter, with a vendor responsible for keeping the machinery current. If the relationship ends, the knowledge stays on campus.

That is the test to apply to any proposal, including ours. Ask what remains in the institution's possession on the day the contract ends. If the answer is nothing, you were renting.