RFP First Draft Generator

Create a consistent folder structure in your document repository (SharePoint, Google Drive, or Notion). Each proposal should be stored as a single document or folder with a standardized naming convention: [Year]-[Client Type]-[Scope Category]-[Outcome]. Create a metadata spreadsheet or database that indexes every proposal in the library with these fields: file path, client industry, scope category, service type, engagement size range, key differentiators used, outcome (won/lost), reason won/lost, methodology language classification (which methodology section does this proposal use), and team credentials referenced. This metadata index is what n8n queries when matching a new RFP to relevant past proposals - without it, the system has to read every proposal from scratch, which is slow and imprecise. Build the index alongside the document library. For each proposal, also extract the reusable content chunks into a separate document: methodology description, firm overview paragraph, relevant case studies (anonymized if necessary), and team bios. Label each chunk with its category and date. These chunks are what the AI assembles into the draft - clean, labeled, separated from the full proposal context.

Create an intake form (using Typeform, Google Forms, or a simple HTML form) with these fields: RFP title, client name, submission deadline, scope description, RFP document upload, and category (from your predefined list). Connect the form to n8n via webhook. When n8n receives the form submission, it downloads the RFP document from the upload URL, extracts the text content, and passes it to an AI node for initial analysis. The AI extraction prompt identifies: the key requirements and scope signals in the RFP, the evaluation criteria (if listed), any mandatory sections, client-specific signals about their priorities or concerns, and the category match from your predefined list. This structured summary is what the system uses to query the wins library.

You are an RFP analyst for a professional services firm. Your job is to read a Request for Proposal and extract the key information needed to find relevant past proposals and begin drafting a response.

Here is the RFP content: {{$json.rfp_text}}

Return ONLY a valid JSON object with these fields:

{
  "rfp_title": "The title or description of this RFP",
  "client_type": "Industry/sector of the issuing organization",
  "scope_summary": "2-3 sentence summary of the core scope of work requested",
  "key_requirements": ["Array of the most important specific requirements"],
  "evaluation_criteria": ["Array of stated evaluation criteria, if listed"],
  "mandatory_sections": ["Array of required sections explicitly stated in the RFP"],
  "client_signals": ["Any signals about client priorities, concerns, or preferences based on how the RFP is written"],
  "scope_category": "The most applicable category from: [list your firm's categories]",
  "engagement_complexity": "simple, moderate, or complex",
  "estimated_response_sections": ["Cover letter", "Firm Overview", "Relevant Experience", "Methodology", "Team Credentials", "Fee Structure"]
}

Using the structured RFP summary, query the metadata index to find the 3 - 5 most relevant past proposals. Match on scope category, client type, and service type. Retrieve the full content of those proposals plus the relevant reusable content chunks. Pass the RFP summary, matched proposals, and content chunks to the AI assembly prompt. The AI maps relevant content from past proposals to the sections of the current RFP, inserts placeholders where custom content is required, and flags sections where the wins library doesn't have strong matches. The assembled draft should follow your standard proposal structure. Every placeholder should be clearly labeled: [CUSTOM: Insert specific differentiator for this client's regulatory context] - not just [INSERT TEXT HERE]. The more specific the placeholder, the easier it is for the reviewer to complete the draft. After draft generation, save the document to a designated folder in your document system and notify the responsible partner with a link to the draft and a summary: what was matched from the wins library, what needs to be written from scratch, and which sections the AI flagged as weak matches.

You are a senior proposal writer for a professional services firm. Your job is to assemble a first-draft RFP response by drawing on past proposals and firm content.

RFP Summary: {{rfp_summary}}

Matched past proposals and content library: {{matched_content}}

Firm's standard proposal structure: {{proposal_structure}}

Instructions:
1. Assemble a complete first-draft response following the standard structure
2. Pull directly from the matched past proposals where content is highly relevant - quote and adapt, don't start from scratch
3. For sections where matched content is only partially relevant, note the gap and write a bridge paragraph that can be customized
4. For sections where no good match exists, insert a clearly labeled placeholder: [CUSTOM REQUIRED: Description of what needs to be written here and what information is needed]
5. Maintain consistent voice throughout - if the tone shifts because you're drawing from multiple proposals, smooth it
6. Flag any factual claims (client names, case outcomes, credentials) that need verification in a "Fact Check Required" section at the end

Format as a properly structured document with clear section headers. This draft should be 70-80% complete - enough that the reviewer's job is refinement and customization, not assembly.

The draft review process needs to include a specific accuracy check - not just a quality review. AI will occasionally generate specific claims that aren't in your wins library. Build a fact-check checklist into your review process: - Every client name mentioned: is it in your approved client reference list? - Every case outcome (dollar amounts, percentage improvements): does it match documented results? - Every credential or certification: does it match current team bios? - Every methodology claim: is it consistent with how your firm actually works? Require the reviewing partner to sign off on factual claims explicitly before the proposal goes out. A one-line addition to your submission email - "Factual claims reviewed and confirmed by [Partner Name]" - creates accountability. After each submission (won or lost), update the wins library. Log the key content used, what was customized, what the client response was, and whether you won. Firms that treat the library as a living document see consistent improvement in draft quality over 6 - 12 months as the library grows.