Fable
The main inquiry desk combines finding materials and organizing answers into one craft
The research institute had many archives: the contract room, the blueprint room, the old case room, and the ledger room. The young assistant could look things up in books and write reports, but every time a question came in, they decided on the spot where to look, how much to take, and how to piece it together. Simple questions could still be handled.
If the director asks "why did the East Bridge repair exceed budget," the assistant either gets only the account books and misses the drawing changes, or brings over a dozen volumes of original text and lets the director search through them. The more material there is, the more the answer feels like a gamble.
The director first decreed that every query must check the first five volumes of the old records room. For account questions, this path was often wrong; for drawing hierarchies and multiple revisions, it was completely inadequate. Each project team later wrote its own lookup cheat sheet, so the same question yielded different evidence at different entry points.
The old curator set up a general inquiry desk. It first breaks down the question: is it asking for facts, comparison, causes, or a summary; then selects the archive room, and if necessary looks for the master volume first, then traces sub-volumes; after retrieving materials it removes duplicates, reorders by importance, and finally hands them to the report writer for integration.
The first version of the general inquiry desk also took too much material, and the reporter was still overwhelmed. The old curator added a retrieval cap and evidence ranking: the most answer-relevant first, background material later, conflicting evidence flagged. Finding it isn't enough; the material must also be organized into something usable.
The main inquiry desk does not itself know all the answers. It knows how to turn a question into a retrieval route, how to gather evidence across multiple sources, and how to control the retrieved material so it is sufficient without drowning the answer.
The research institute later discovered that what truly made Q&A stable was not some assistant being clever on the spot, but a desk that could organize queries, select sources, rerank evidence, and deliver materials. Without it, the more data there was, the easier it was for answers to get lost.
Later, the main inquiry desk also learned to refuse. If a question was too broad, it would first ask the director to narrow the scope; if materials conflicted, it would hand over the conflict instead of forcing it into a smooth report. As a result, the institute had much less beautiful nonsense and more answers that could be traced back to their sources. Later, the assistants also felt relieved, because they no longer had to guess from scratch which archive room to go to every time.