The prompt looks simple. The assumptions don't.
The organization is doing the hard part silently.
"Show me MRR by segment for last quarter." Seven words. Six unresolved questions hidden below the surface, each one representing a decision the business has already made but never made explicit.
Which accounts qualify as MRR? An analyst knows to exclude trials, internal, migrated, and one-time contracts. The prompt doesn't say. What is 'last quarter'? Fiscal Q or calendar Q? Which timezone? Closed-won date or invoiced date? Again: the analyst knows. The prompt doesn't say.
Which segment definition? Engineering uses company size. Sales uses ARR band. Finance uses industry vertical. Which MRR table? mrr_monthly, mrr_recognized, and mrr_contracted all exist, and they produce different numbers.
The prompt feels simple because experienced analysts have internalized all of these resolutions. They apply them without thinking. That implicit knowledge looks like speed. For a system with no access to it, it looks like ambiguity.
Every organization has a layer of operational context that makes simple questions answerable. That layer isn't in the schema, the catalog, or the prompt. Until a system can read it, every simple query is a guess.
