Hiring AlphaCorp AI for custom RAG: can your team rerun the checks?

Use a fictional manual, expected answers and source references to decide whether to outsource read-only document Q&A.

Source: alphacorp.ai
August 26, 2026
2 min

AlphaCorp AI offers custom retrieval-augmented generation (RAG): its service page lists a chat interface or API with citations and feedback. That is not a ready-made document search app. For a few dozen pages where staff only need the original passage, a versioned index and browser search may suffice. Compare an in-house build with outsourcing only when finding passages across manuals and checking their versions takes appreciable time.

Article img
Two P-02 versions and two questions go into the pack. The v2 retest points T-01 to paragraph 2 and leaves T-02 unsupported.

Figure 1. Original handover diagram, not a service test. A second maintainer should be able to rerun the same pack.

Agree on a handover test before the demo

Write two fictional equipment-labelling manuals named P-02. T-01 asks where the label goes now and must cite paragraph 2 of v2; v1 differs. T-02 asks for a label colour the manual never gives; the answer must say there is no source. Request a test pack independent of the vendor's demo: the documents and version list, questions, expected output and rerun instructions. Ask the receiving maintainer to rerun it in an agreed isolated environment. Change a paragraph number without changing the rule, then verify that the maintainer can refresh the index, find the stale citation and record the rerun. Use no real systems, private documents or credentials.

Ask who owns initial delivery, reindexing and regression investigation, and how each is priced. Confirm whether the pack, configuration and exportable results are deliverables. Consider a small engagement only if both questions and the version-change rerun pass with a second maintainer. If only the vendor can demonstrate that it works, stick with the versioned index or build a read-only prototype with your own maintainer.

Content note:This article is based on publicly available information and is intended as a reference for evaluating tools. Features, pricing, and related policies may change at any time; please refer to the latest official information.

Original source:alphacorp.ai