Platform team weighing whether to build a generative AI capability in-house or license from a vendor pitching a hybrid retrieval / chat product. Concerns: defensibility, data and IP exposure, evaluation pipeline gaps, retrieval architecture quality (chunking, embedding, hybrid fusion, reranking, query rewriting / routing / decomposition), local-LLM viability, sovereignty.
Technical advisor mandate. Independent assessment of the vendor's pitch against modern retrieval-augmented architecture practice. Build-vs-buy analysis grounded in long-term cost, future-proofing, evaluation discipline and data-pipeline ownership. Concrete written critique covering each architectural concern with the trade-offs spelled out.
Strategic clarity for the platform team on which path actually compounded. Internal capability path articulated end-to-end (data pipeline → embedding → hybrid retrieval → eval pipelines → on-prem viability), with the licensing path properly sized rather than dismissed.
One call to agree whether a process audit, a system build or a fractional AI lead fits where you are.