# How to Respond to an RFP with Claude: Go/No-Go to First Draft
Learn how to respond to an RFP with Claude, step by step: project setup, go/no-go, a compliance matrix, win themes, and a first draft that flags gaps.
Most teams open the RFP and start writing. That puts the blank document before the decisions that should shape it.
This walkthrough takes one mock opportunity through a focused [RFP response process](/blog/rfp-response-process) in Claude. It is a single respond path, from project setup and qualification to a first draft. If you want the broader proposal craft and a catalogue of reusable Claude skills, use the companion guide on [how to write a bid proposal](/blog/how-to-write-a-bid-proposal).
Claude is useful here because it can hold a long RFP, follow saved instructions, and produce structured working documents. Your team still owns the bid decision, strategy, commercial judgment, and final review.
AutoRFP.ai takes an accuracy-first approach to this work. It drafts from content your team has approved, shows the sources behind each answer, and routes unsupported questions to a person. That is the standard to keep in mind as you build a manual Claude workflow.
## Set up a Claude project before you write
Start by creating one Claude project for the opportunity. In the video, I use a fake company and a fake RFP so the full workflow can be shown without exposing a live bid.
Add the context Claude will need:
- Project instructions that define the job, the order of work, and the expected output.
- Past RFP responses, labelled so Claude can tell which document and section it is using.
- Security documents, policies, technical material, pricing guidance, and company information.
- Examples of answers your team considers strong.
- Notes from earlier conversations with the prospect, where they are approved for bid use.
The examples matter. Company facts tell Claude what is true. Past answers also show how your team explains those facts.
Next, ask Claude to create a file called `project context map.md`. Its job is to list the files in the project, what each contains, and when Claude should use it. Point the project instructions to that map. This gives Claude a repeatable route to the relevant context instead of leaving it to inspect every file from scratch each time.
The video uses Claude skills for the specialist jobs. A skill is a saved set of instructions, supporting files, and frameworks. For example, the [AI Go/No-Go Agent Skill](/skills/ai-go-no-go-agent-skill) carries the qualification steps and scorecard so the team does not rebuild a complicated prompt for every opportunity. Customize any skill to your own criteria before sharing it across the team.
The Cowork project instructions used in the recording are available from the video description. Treat them as a starting point, then change the language, decision rules, and source hierarchy to match your company.
## Run an AI go/no-go before a word of the draft
A go/no-go asks whether the opportunity deserves your team's bid effort. It should happen before writers begin filling pages.
The wider [Go/No-Go framework for RFPs](/blog/rfp-go-no-go) covers strategic fit, delivery capability, buyer relationship, pricing, references, competitive position, and the resources needed to submit. Claude can collect evidence and apply the scorecard, but it cannot make the final business decision for you.
> **71%**: 71% of high-win teams have a Go/No-Go qualification step (72% of low-win teams also report one). (Source: Proposal Win Rate Report 2026, 9 Dec 2025)
In the mock example, Claude returns a Conditional Go at 77 out of 100. It explains the result through the scorecard, including the relevance of named references, pricing information, prior commitments, and possible contradictions. That explanation is more useful than the score alone because it gives the bid team specific points to challenge.
Bring the output to the people who own the deal. Confirm the assumptions, supply any missing information, and decide whether to proceed. A well formatted AI report does not remove accountability from the team.
## Build the compliance matrix and the submission checklist
Once the team decides to bid, extract the rules before drafting the response.
The [Compliance Matrix Builder Agent Skill](/skills/compliance-matrix-builder-agent-skill) used in the video turns the RFP into four working views:
1. A requirements matrix that preserves what the buyer asked.
2. A submission checklist with actions and dates.
3. The evaluation criteria and scoring structure.
4. A mock evaluation showing where the current bid may need attention.
This is the practical purpose of an [RFP compliance matrix](/blog/compliance-matrix). It connects each requirement to an owner and a place in the response, while keeping submission instructions visible.
Compliance and responsiveness are different jobs. The matrix helps you satisfy the instructions that keep the proposal eligible. The evaluation criteria tell you what the buyer is likely to reward. A response can tick every box and still score poorly if it does not address the buyer's priorities, risks, and desired outcomes.
Use the checklist to assign attachments, approvals, and deadlines. Use the criteria to decide how much space and proof each section deserves. Only then should drafting begin.
## Generate win themes, then map them to sections
A win theme gives the evaluator a clear, supportable reason to choose you. It connects something the buyer cares about with a strength you can prove.
Before generating themes, give Claude the context behind the opportunity:
- Why the buyer is acting now.
- Which stakeholders will evaluate or live with the solution.
- What each stakeholder cares about.
- Who else may be competing and where your evidence is genuinely different.
- Which evaluation criteria carry the decision.
In the mock scenario, the prospect is a talent acquisition company with a team of two planning to hire 90 people. Claude identifies themes around creating capacity, consolidating hiring, and producing board-ready reporting. It then maps each theme to the response sections where it should appear.
The important step is the mapping. A list of slogans at the top of the project will not influence the evaluation. The themes need to shape the executive summary, solution sections, implementation approach, and proof. The full guide to [win themes](/blog/win-themes) explains how to test each one against buyer priorities and available evidence.
Ask the deal team to challenge the themes before drafting. Remove any that a competitor could claim without changing a word. Strengthen any theme that lacks a case study, reference, method, or other approved proof.
## Draft in the buyer's file and flag the gaps
Now ask Claude to draft against the matrix, the agreed themes, and the company context. Keep the output in the buyer's structure. If the RFP arrived in a Word or spreadsheet format, preserve its order and response fields so the first draft remains easy to review and return.
The recording shows Claude producing an internal draft, including an [RFP executive summary](/blog/rfp-executive-summary), then marking areas where it cannot find enough context. Those gaps are useful. They tell the proposal manager what needs a subject matter expert, a commercial owner, or more approved evidence.
Use a clear instruction such as:
> Write only from the project files. If the sources do not support an answer, mark the requirement as a gap, explain what is missing, and assign the question for human review. Do not invent a customer fact, metric, reference, price, or product capability.
Reviewers should be able to see the difference between sourced draft text and an unresolved requirement. Do not let polished prose hide uncertainty.
Move the working draft into Google Docs or SharePoint when the wider team needs to contribute. Assign each gap to the right person. After the response is approved and submitted, save the completed document back into the project as last year's answers. Label it clearly so Claude can use the approved result on the next relevant bid.
This creates a useful loop: every finished response improves the examples and facts available to the next project. Someone still needs to own freshness, remove superseded information, and make sure Claude is drawing from the right version.
## Where a shared team outgrows the Claude project
Claude is a strong workspace for learning this method and proving the sequence on one opportunity. It can analyze long documents, follow reusable skills, and get a team to a structured first draft.
The limits become clearer when several people are responding across multiple bids. Files live in different systems. Owners need assignments and deadlines. Reviewers need to know which source supports an answer without asking the model to explain itself after generation. The final response also has to return in the buyer's format.
This is where purpose-built [AI proposal software](/ai-proposal-software) replaces a collection of manual handoffs. AutoRFP.ai drafts from approved sources, shows the evidence behind each answer, and routes anything it cannot support to a person. The team can collaborate on the response while keeping the review trail attached to the work.
The practical sequence stays the same:
1. Load the right company and opportunity context.
2. Decide whether to bid.
3. Map the requirements and evaluation criteria.
4. Agree on win themes.
5. Draft from approved evidence and hand the gaps to people.
Claude can help you learn and run that sequence. The quality of the result still depends on the process around it and the evidence you provide.