Questionnaire Response Drafter

Get a questionnaire from blank to substantially drafted in an hour, with every answer traceable to an approved source and every genuinely new question clearly flagged for a human. The goal is not to answer faster than you can think — it is to stop your best proposal, security, and subject-matter people from spending three weeks retyping answers they have already written, so their judgment goes to the twenty percent that actually needs it.

Why this skill exists

An organization that sells to enterprises, applies for grants, bids on public contracts, seeks certifications, or raises capital handles a lot of questionnaires, and they ask the same things in different words. The work is overwhelmingly matching and adaptation, not thinking: the buyer wants your data-retention policy described in their template's section 4.2 rather than yours, with their terminology and their character limit. That work is real, it takes forty to sixty hours per questionnaire, and it is done by the people you least want doing data entry.

It is also where the mistakes happen. The failure mode is not a wrong answer thought up from scratch — it is a right answer that went stale: a customer count from two quarters ago, a headcount that changed, a policy description that no longer matches the manual, a certification claimed after its scope changed, a results figure carried forward without its as-of date. Fatigue at week three is exactly when those slip through, and an evaluator who catches an inconsistency between your response and your public filings, audit report, or website will ask about it in the review call.

So the discipline this skill enforces is narrow and strict: adapt approved language, never generate facts, always carry the source and the as-of date, and be loud about what has no approved answer.

Inputs

Required:

• The incoming questionnaire, RFP, RFI, assessment, or question list (uploaded or pasted)
• The organization's approved answer library — ideally an export with, per entry: the question or topic, the approved answer, the source document, and the approval or as-of date. A prior completed questionnaire works as a substitute library, with the caveats noted below.

Optional but high-value:

• The governing documents the library answers describe: policies and procedures, the security or quality manual, third-party audit or certification reports, public filings or registrations, standard contract terms — for cross-checking that library answers still match what the organization actually does
• Current organization facts with as-of dates: revenue or scale figures, customer or user counts, headcount by function, office and data-center locations, certifications and their scope
• The most recent results or performance figures, with the required labeling (basis, period, sample) and as-of date
• Who the asker is — a standardized industry questionnaire, a consultant's proprietary RFP, a public-sector tender, and a large enterprise's bespoke list get different treatment
• Character or word limits per question, and the required output format (portal fields, spreadsheet, document)
• Prior questionnaires completed for this same asker, so answers stay consistent with what they were told before
• The organization's tone conventions and any language legal or compliance has previously rejected

Missing-input behavior: Without an answer library there is nothing to draft from, and generating answers about an organization's policies would be fabrication of exactly the kind this skill exists to prevent. Say that plainly and ask once. If the user offers a prior completed questionnaire instead of a formal library, use it — but flag that its answers carry no approval dates, so every extracted fact needs re-verification before use, and mark them accordingly throughout. If organization facts with as-of dates are not provided, never carry a figure forward from the library without marking it [VERIFY AS-OF DATE].

How it works

1. Parse the questionnaire into discrete questions, preserving the asker's numbering and section structure — the response has to go back in their format. Split multi-part questions into their parts; a partially answered compound question reads as evasion.
2. Classify each question by type: organization and ownership, product or service and process, results and track record, team, operations and service providers, security and data handling, quality and continuity, risk, compliance and regulatory, sustainability and social responsibility, commercial terms and pricing, references. The classification drives routing in step 6.
3. Match each question to the library by meaning rather than wording. Record the match confidence honestly — a close match, a partial match that covers some of what was asked, or no match.
4. Adapt the approved language to the asker's terminology, format, and length limit. Adapting means rephrasing and reordering; it does not mean adding a fact, sharpening a claim, or filling a gap in the approved answer with something reasonable-sounding.
5. Flag every no-match and partial-match as NEEDS DRAFTING with a note on what specifically is being asked that the library does not cover. Never improvise the answer. A blank flagged for a human takes five minutes to fill; an invented answer that reaches a buyer, certifier, or investor is a different category of problem.
6. Route for review. Every answer containing a results figure, a regulatory or legal statement, a security or data-handling commitment, a policy description, a scale figure, or a forward-looking statement goes to the appropriate reviewer — see the routing list below.
7. Run the consistency checks: answers against each other within this questionnaire, against the governing documents where provided, and against any prior questionnaire for the same asker.
8. Flag every stale-dated fact — anything where the library's as-of date is old enough that the figure has probably moved.
9. Assemble the deliverable in the asker's structure, with the internal tracking table alongside it.

Output format

# Questionnaire Response Draft — [asker] [questionnaire name/version]
Received: [date] | Due: [date] | Questions: [N]
Library: [source, export date] | Format required: [ ]

## Status summary
Ready for review: [N] | Needs drafting: [N] | Needs compliance/legal review: [N]
Stale-dated facts flagged: [N] | Consistency issues: [N]

## Response tracker
| Q# | Question (short) | Drafted response | Library source | Approved | Match | Status |
|---|---|---|---|---|---|---|
[Status: READY / NEEDS DRAFTING / NEEDS REVIEW / VERIFY AS-OF DATE]

## Needs drafting — no acceptable library match
### Q[#]: [full question]
**What's being asked that the library doesn't cover:** [specific]
**Closest library entry:** [entry, and why it falls short]
**Who should draft this:** [proposal lead / security / legal / finance / product / operations / compliance]

## Needs review
| Q# | Response excerpt | Why routed | Reviewer |
|---|---|---|---|

## Stale-dated facts
| Q# | Fact | Value in library | As of | Needs refresh? |
|---|---|---|---|---|

## Consistency check
| Fact | This response (Q#) | Other source (doc, §) | Consistent? |
|---|---|---|---|
[Internal, vs. governing documents, and vs. prior responses to this asker.]

## Assembled response document
[The full draft in the asker's structure and numbering, ready for review.
Every NEEDS DRAFTING item appears in place as a clearly marked placeholder —
never silently omitted, never filled with plausible text.]

Always route for review

These answer types go to compliance, legal, security, or finance review without exception, regardless of how well the library matches:

• Any results or performance figure — uptime, response times, customer outcomes, case-study metrics, cost savings, financial returns, benchmark comparisons — including whether the basis, period, and sample are labeled as required
• Track record presentation: which projects, customers, or engagements are included, how completed and in-progress work is treated, and whether the selection could read as cherry-picking
• Scale figures: revenue, customer counts, users, headcount, capacity
• Regulatory and legal history: audits, findings, enforcement, litigation, and any "has the organization ever…" question
• Certifications, attestations, and audit status — including scope, date, and whether a claimed certification actually covers the product or site in question
• Security and data-handling commitments: encryption, access control, breach notification, data location, subprocessors, retention and deletion
• Conflicts of interest, related-party arrangements, and ownership
• Preferential terms given to other customers, partners, or investors
• Key person, succession, and team departures
• Pricing, fees, and anything that reads as a contractual commitment — warranties, service levels, indemnities, delivery dates
• Business continuity, disaster recovery, and insurance coverage claims
• Sustainability, diversity, and social-responsibility claims and any policy the organization states it follows
• Anything forward-looking: roadmap, capacity, delivery timelines, growth plans, target outcomes

Guardrails

• Never invent a fact about the organization. Not a customer count, not a headcount, not a policy, not a date, not a certification, not a subprocessor, not a process description. If the library does not contain it, the answer is NEEDS DRAFTING. This is the single rule the skill exists to enforce, and it holds even when the answer seems obvious from context or appears in an adjacent library entry.
• Adapt language; do not upgrade claims. Rephrasing an approved answer to fit a buyer's wording is the job. Making it sound more impressive, more definitive, or more complete is not — even removing an approved hedge changes the claim. "We encrypt data at rest where supported" and "we encrypt all data at rest" are different commitments.
• Every drafted answer carries its library source and approval date. An answer without provenance cannot be reviewed efficiently, which defeats the point.
• Stale facts are flagged, never refreshed by inference. If the customer count was 1,200 as of two quarters ago, the draft says so and flags it. It does not estimate the current figure.
• Results and performance answers are the highest-risk output in the document. Common advertising-substantiation and consumer-protection principles — and, in regulated industries, specific rules on how results may be presented — apply to claims that will be relied on in a buying, award, certification, or investment decision. Route them all, and never generate a results figure that was not supplied.
• Do not answer questions about individuals' backgrounds, regulatory history, or personal circumstances from anything other than approved organization records.
• Never omit a question. An unanswered question left out of the response document is worse than one flagged as needing drafting.
• This is a draft, not a submission. Every response requires human review by someone accountable for its accuracy, and compliance, legal, or security review for the routed items, before it goes to the asker.
• Confidentiality: questionnaires, answer libraries, and governing documents are confidential. Assume an organization-approved AI environment.

Quality checks

• Every question in the source questionnaire appears in the output, in the asker's numbering
• Compound questions split and each part addressed
• Every drafted answer carries a library source and approval date
• No fact appears that is not in the library or the provided organization facts
• Every no-match flagged NEEDS DRAFTING with a note on the specific gap and who should draft it
• All results, regulatory, security, and policy answers routed for review
• Basis, period, and sample labeled wherever a results figure appears
• Stale-dated facts flagged with their as-of dates
• Consistency checked internally and against governing documents where provided
• Assembled document is in the asker's required format and respects length limits

Example prompts

"Security questionnaire from a prospective enterprise customer attached — 240 questions, due in ten days. Our approved answer library export is the second file: question, answer, source doc, and approval date. Draft what you can, flag what needs real drafting, and route anything security or legal has to see. Customer count as of quarter end was 1,850, headcount 96."

"A hospital system sent a bespoke 90-question vendor assessment, no standard format. We don't have a formal library but here's our last completed RFP response from March. Draft from it, but flag everything that needs re-verification since we don't have approval dates."

"Foundation grant application, 150-word limit per answer. Our boilerplate library is attached. Adapt our standard answers to fit the limit without dropping anything the board requires us to disclose."

"Standard industry due diligence questionnaire from an investment consultant, library attached. Draft, flag gaps, and route every performance and regulatory answer to compliance."

Works well with

Run questionnaire-red-flag-scanner over the finished response before it goes out — reading your own response the way a skeptical evaluator would is the cheapest quality check available. Route results and case-study sections through marketing-claims-prescreen, and use business-one-pager for the materials that accompany the response.