A response desk for RFPs and security questionnaires

Build me a questionnaire response desk that my security and sales teams work out of when an inbound RFP, security questionnaire, or vendor due diligence form arrives. Today we copy paste answers out of old spreadsheets, which is slow and gives inconsistent wording. I want one screen where every question in the questionnaire has a drafted answer, the approved source document that answer came from, and a status a reviewer can move.

When I open the app I pick the questionnaire I am working on, or set up a new one. Setup takes the Google Sheets spreadsheet ID of the questionnaire, the tab and range where the questions live, the column to write approved answers back into, the Google Drive folder ID of our approved source material, and the spreadsheet ID of our answer library sheet. Save these per questionnaire so several can be in flight at once and I can come back to any of them.

An import handler reads every question using Google Sheets Get Values for a single range, and Batch Get Values when the questionnaire spans several tabs or column ranges. Questionnaires arrive in different shapes, so before the import runs let me map which column holds the question, which holds the answer, and which holds any section or category label. Each imported question becomes a row in the desk.

For each question, a drafting handler finds relevant approved source material in Google Drive with List Files scoped to the approved folder, then pulls the document content: Export Google Workspace File to PDF for Google Docs, Sheets, and Slides, and Download File Content for PDFs and other uploaded files. It passes those documents straight into AI Generation Generate Completion as file inputs rather than pre-summarizing them, and asks for a drafted answer, the specific document and section the answer is based on, and a confidence rating. Before drafting from scratch, check the answer library sheet with Get Values for an already approved answer to a near identical question and reuse that wording when there is a strong match.

The main view is a question by question review table. Each row shows the question, the drafted answer, the source document it came from with a link to open it in Drive, the confidence rating, and a status of draft, approved, or needs expert review. Selecting a row opens a detail panel with the full draft, the passages it was drawn from, and three actions: approve as written, rewrite and approve, or escalate to an expert with a note. Reviewers can filter by status and by section so two people can split a long questionnaire. Persist statuses, edited wording, reviewer notes, and escalations in the app so work in progress survives a reload and a second reviewer sees the same state.

Low confidence answers must be flagged, never silently filled in. If the model is not confident, or no approved document actually covers the question, the row lands as needs expert review with an empty answer and a short note saying what is missing. A wrong security answer is worse than a blank one, so nothing is written back to the questionnaire until a person approves it.

Approving writes the answer into the original questionnaire spreadsheet with Update Values for a single row, and Batch Update Values when a reviewer approves several rows at once. Every approved answer is also appended to the reusable answer library sheet with Append Values, recording the question, the approved wording, the source document, the questionnaire it came from, who approved it, and the date. Show a progress summary across the top of the desk: total questions, drafted, approved, and awaiting expert review.

Give me a second view for the answer library itself, read with Get Values, so the team can browse and search everything we have already approved and see which answers keep getting reused. This library is the point of the app: each questionnaire we finish should make the next one faster because we start from our own approved wording instead of a blank sheet.