RFPs are the most expensive sales activity in any agency. A single response typically eats 20-40 hours of senior time. Multiply that across 3-5 active RFPs per month and you have a full-time employee’s worth of unbillable hours going into documents that have a 20-30% win rate at best.
Most agencies either grind through every RFP that comes in (burning out their best people) or avoid them entirely (missing opportunities worth six figures). AI offers a third option: respond to more RFPs, in less time, with better answers.
Why RFPs eat so much time
Break down where the hours actually go and a pattern emerges.
- Reading and understanding the brief: 2-4 hours. RFPs are long, often poorly structured, and full of requirements buried in appendices.
- Research and strategic thinking: 4-6 hours. Understanding the client’s market, their current position, and what they actually need (which is rarely what the RFP says they need).
- Writing responses to individual questions: 8-15 hours. This is the bulk. Dozens of questions, each requiring a tailored answer that demonstrates understanding without being generic.
- Case study selection and writing: 3-5 hours. Choosing the right examples and rewriting them to address the specific evaluation criteria.
- Formatting, proofing, and polishing: 3-5 hours. Getting the document to a standard that reflects your agency’s quality.
AI can substantially reduce every single one of these stages. Not eliminate them. Reduce them. The total time drops from 20-40 hours to 8-15 hours, with better output quality.
Building your RFP knowledge base
Before AI can help with RFPs, it needs material to work from. This is the step most agencies skip, and it is the one that makes everything else work.
Your knowledge base should include:
- Previous winning responses. Every RFP you have won, archived with the questions and your answers. Tag each answer by theme: methodology, team structure, measurement, pricing approach.
- Case studies in long and short form. Full case studies (500-800 words) and summaries (100-150 words) for each major project. Include metrics and specific outcomes.
- Standard boilerplate that is actually good. Your agency overview, team bios, process descriptions, data security policies. These appear in almost every RFP. Write them well once.
- Past briefs and your strategic responses. The “our approach” sections from previous pitches.
Store this in a format AI can access easily: a structured document, a Notion database, or a folder of clearly named files you can paste into Claude or ChatGPT as context.
This knowledge base pays dividends beyond RFPs. It feeds into your proposal writing process, your pitch decks, and your general new business pipeline.
The AI-assisted RFP workflow
Here is how the process works once your knowledge base is in place.
Step 1: Brief analysis (30 minutes instead of 3 hours).
Feed the entire RFP document into Claude. Getting the most out of this step depends on effective prompt engineering. Ask it to extract: the key requirements, evaluation criteria and their weightings, mandatory compliance points, submission deadlines and format requirements, and any ambiguous sections that need clarification.
Then ask it to compare this brief against your knowledge base and flag which previous responses are most relevant. This gives you a head start on the strategic approach and immediately surfaces gaps where you will need to write something new.
Step 2: Draft responses from previous wins (2-3 hours instead of 10).
For each question in the RFP, ask AI to draft a response using relevant content from your knowledge base. The prompt matters here. Do not just say “answer this question.” Say: “Draft a response to this question for a [type of client] in the [industry] sector. Draw on the following previous responses [paste relevant past answers]. Tailor the language to address their specific evaluation criteria of [criteria]. Our key differentiator for this pitch is [differentiator].”
The output is a first draft, not a finished answer. But it is a first draft that is already 70-80% there, drawing on your best previous work rather than starting from a blank page.
Step 3: Tailoring to the brief (2-3 hours).
This is where senior input matters. Review each drafted response and ask: does this actually answer what they are asking? Are we referencing their specific challenges? Is the case study genuinely the best fit? Does the tone match their organisation? (A government body reads differently to a startup.)
AI helps here too. Feed it the prospect’s annual report and website. Ask it to identify specific language, priorities, and themes you should mirror. This is the same competitive analysis approach applied to the RFP process.
Step 4: Case study rewriting (1 hour instead of 4).
Select case studies based on relevance to the evaluation criteria, not just recency or impressiveness. Then use AI to rewrite each one, emphasising the aspects that map to what this specific prospect cares about.
If the evaluation criteria weight “measurement and ROI” heavily, your case study should lead with the numbers. If they weight “creative approach,” lead with the strategic thinking and creative execution. Same project, different framing.
Step 5: Quality control and polish (1-2 hours).
AI drafts well but it does not have taste. The final review is where a senior person ensures the response sounds like your agency, not like a competent but generic document.
Check for: consistency of tone, strategic coherence (does the whole response tell one story?), specificity (replace any vague language with concrete examples), and compliance (have you actually answered every question and met every requirement?).
What still needs senior input
AI handles the assembly and drafting. Humans handle the strategy.
The “why us” argument. Every RFP response needs a thread that explains why your agency, specifically, is the right choice. This requires self-awareness about your genuine strengths and an honest read on what the client values. AI cannot do this for you.
Pricing strategy. How you price, how you present your rates, whether you offer phased options, where you build in contingency. These are strategic decisions informed by experience and commercial instinct.
Reading between the lines. RFPs rarely say exactly what the client wants. “We are looking for a creative agency partner” might mean “our current agency is too expensive” or “our board wants to see innovation” or “the marketing director needs to justify their budget.” Experienced agency people read these signals. AI does not.
The go/no-go decision. Not every RFP is worth responding to. The decision to invest 10-15 hours of team time should be based on realistic win probability, strategic fit, and commercial value. AI can help you score the opportunity, but the final call is yours.
Measuring the impact
Track four numbers: hours per response (expect a 40-60% reduction), responses submitted per month (you can now pursue more), win rate (better tailoring should push this up by 10-20 percentage points), and cost per pursuit (fully loaded senior time per response).
The maths is simple. If you currently spend 30 hours on each RFP at an average senior rate of £120/hour, each response costs £3,600. Cut that to 12 hours and it costs £1,440. Respond to 5 RFPs per month and you save £10,800 monthly, or nearly £130,000 per year. That is before accounting for the improved win rate.
This is one of the clearest ROI cases for AI in the entire agency sales process. Start building your knowledge base today. The first RFP you respond to with AI support will convince you.
This is part of The Pitch Stack, a series on agency sales and new business strategy. Subscribe to the newsletter to get new articles weekly.
