What is the model?
Illustrative, with every assumption stated. Fifteen tenders a year; six hours of writing per tender retrievable from approved content at a $180 blend: $16,200 gross annual saving if every retrieval fits perfectly, which none do. Against it: adaptation (2 hours per reuse), verification against the live pack (1 hour per reuse, non-negotiable, skipped at peril), and upkeep (4 hours a month curating, retiring, and re-verifying content): roughly $16,200 in the same arithmetic. Net at these assumptions: approximately zero, before counting faster lodgement, calmer review weeks, and fewer errors: real benefits, harder to price, honestly labelled as unmodelled.
Change the reuse rate and the answer moves: high-reuse firms with stable scopes go positive; low-reuse firms with bespoke scopes go negative. That is the entire point; run your numbers, not anyone’s benchmark. Response tooling that hosts the library is a separate decision owned by AI bid-writing tools; close-out extraction that feeds it is on cost of a lost bid; spreadsheet discipline around it is on spreadsheets and SharePoint.
What does this mean for the bid/no-bid call?
Nothing directly; a library makes responses cheaper, not decisions better, and cheaper responses to poor-fit tenders are how firms lose money efficiently. Fund the library from response economics; fund the gate from pursuit economics; never let the first justify weakening the second.