Turn a Block of Client Meeting Notes into a Structured RFP Response Outline

The task

You just wrapped a discovery call with a prospect. You have a wall of typed notes — quotes, side comments, half-formed requirements — and a proposal due in 48 hours. This workflow turns that raw block into a structured RFP response outline you can hand to the pursuit team.

Before AI

Most consultants do this by hand: re-read the notes, highlight in three colors, transcribe into a Word skeleton, then argue about section ordering. On a complex transformation deal — think supply chain AI, ERP replacement, cloud migration — that's easily two hours before anyone writes actual proposal content. Miss a stated requirement and you get scored down in evaluation.

The signal driving urgency right now: multi-agent AI systems are taking over supply chain execution, and clients are asking vendors for rapid deployment proposals with tight turnarounds. Speed on the outline stage is where hours get won back. For the underlying structure, this workflow leans on the standard RFP response format used across consulting bids.

The workflow

Step 1 — Extract structured requirements from the notes

Paste your raw notes at the end of this prompt. It pulls out what the client actually said they need, separated from context and noise.

You are a senior consulting proposal lead. I will paste raw discovery-call notes below. Your job is to extract structured information ONLY from what appears in the notes — do not invent requirements, budgets, timelines, or stakeholders that are not stated or clearly implied.

Return your output as the following labeled sections in plain markdown:

1. **Client context** — company, industry, current state, business trigger (2-4 bullets)
2. **Stated requirements** — explicit "we need / we want / must have" items, quoted or paraphrased tightly (bulleted)
3. **Implied requirements** — things a competent bidder would infer from the conversation but that the client did not say outright (bulleted, each marked "(inferred)")
4. **Stakeholders mentioned** — name, role, and what they care about
5. **Constraints** — budget, timeline, tech, compliance, political
6. **Open questions** — things you'd want clarified before writing the proposal
7. **Risk flags** — anything in the notes that suggests scope creep, unrealistic expectations, or a competitor already in the door

If a section has no content from the notes, write "None stated" — do not fill it in.

Raw notes follow:
---

Discovery call — Northwind Foods, Sept 22. Present: Priya Ramaswamy (VP Supply Chain), Dan O'Neill (Head of IT), Marcus Chen (CFO, dropped off after 20 min).

Priya opened by saying their current dashboards from LogiView are "basically wallpaper" — nobody looks at them because by the time an exception shows up the truck's already late. They want something that actually acts, not just reports. She mentioned a competitor (didn't name) is running "agent-based" routing and cut their on-time misses by a third.

Northwind: ~$2.1B revenue, 14 DCs across US and Mexico, SAP ECC (upgrading to S/4 next year — Dan was firm this project cannot block that). Peak season is Oct-Dec, they want pilot value before Q4 next year.

Dan pushed hard on data residency — Mexico DCs cannot send order data outside Mexico, apparently a new interpretation from their legal team. Also asked twice about how we handle "hallucinations" in agent decisions. He seems burned by a prior GenAI POC that went nowhere.

Marcus (before dropping): budget "meaningful but not open-ended," mentioned $3-5M range for year one including our fees. Wants a business case tied to working capital reduction, not just OpEp savings.

Priya's must-haves: exception handling on inbound freight, dynamic reslotting in the DCs, and — this was interesting — she wants ops managers to be able to override the agents and have the system learn from the override. Said "if my people feel replaced, this dies."

Timeline: RFP response due Oct 6. Shortlist decision end of October. Wants pilot live by March. Two other firms bidding — she named one, Accenture. Wouldn't say the other.

Random: Dan mentioned they already have Databricks. Priya used the phrase "multi-agent" twice unprompted. Nobody mentioned change management budget.

Step 2 — Map requirements to a proposal outline

This step takes the extracted structure and builds the actual RFP response skeleton, with each section pre-loaded with the requirements it needs to address.

Using the structured extraction above, generate a proposal outline for the RFP response. Use this standard consulting RFP structure:

1. Executive Summary
2. Understanding of Client Situation
3. Proposed Solution / Approach
4. Delivery Methodology & Phasing
5. Team & Governance
6. Timeline & Milestones
7. Commercials & Assumptions
8. Risks & Mitigations
9. Why Us / Differentiators
10. Appendices (case studies, references, security posture)

For EACH section:
- List the specific requirements, stakeholder concerns, and constraints from the extraction that this section must address. Reference them by short tag (e.g. "Data residency — MX DCs").
- Write 2-3 bullet points describing what the section should argue or prove.
- Flag any section where the notes gave you nothing to work with — mark it "NEEDS INPUT: [what to ask]" rather than inventing content.

Do not draft prose yet. This is a working outline for the pursuit team.

Step 3 — Generate the clarification email

Before anyone writes proposal content, you send the open questions back to the client. This drafts that email using only what was flagged in Step 1.

Draft a short, professional email to the client's primary contact from the discovery call. The email should:

- Thank them for the conversation (one line, no gushing)
- Confirm you're proceeding with the response
- Ask ONLY the open questions and "NEEDS INPUT" items surfaced in the previous steps — grouped logically, maximum 6 questions
- Note any assumptions you'll make if they don't respond in time, tied to the RFP due date from the notes
- Close with a specific next step

Tone: peer-to-peer, senior consultant to senior client. No "I hope this email finds you well." No filler.

Gotchas

• The model will invent requirements if you let it. The "do not invent" instruction in Step 1 matters — check the "Implied requirements" section for anything that isn't actually supported by the notes. Delete aggressively.
• Quotes get paraphrased. If you plan to use a client quote verbatim in the Executive Summary, pull it from your original notes, not from the model's output.
• Political dynamics get flattened. The model catches "Priya said X" but misses "Priya said X while Dan looked annoyed." If your notes captured tone, add a manual pass before Step 2.
• Compliance items need a human. Data residency, security posture, and regulated-industry constraints (HIPAA, GDPR, ITAR) should be verified by your firm's compliance lead before they land in the proposal — the model will happily write confident sentences about controls you don't have.
• Don't skip Step 3. The open questions list is the most valuable output. Sending it fast signals rigor and buys you time.

Time saved

Manual outlining from a discovery-call dump runs 90 minutes to two hours for a mid-complexity deal, longer if two people argue about section order. This workflow gets you to a working outline and a clarification email in about 15 minutes of prompt-and-review. Call it 60-90 minutes saved per proposal, more on complex transformation bids. The savings are in the mechanical extraction and structuring — the strategic thinking about win themes still happens in your head.

What we got when we ran it

We executed every prompt above, in order, against the sample input (model: claude-sonnet-4-6). Outputs below are real and unedited apart from length truncation.

Step 1 output

# Northwind Foods — Discovery Call Extraction Call date: September 22 | Our note-taker's deadline: RFP response due October 6 --- ## 1. Client Context - Company: Northwind Foods — ~$2.1B revenue, 14 distribution centers across the US and Mexico; industry: food/CPG supply chain - Current state: Using LogiView for supply chain dashboards; client perception is that alerts surface too late to be actionable ("basically wallpaper") — exception visibility exists but no prescriptive or automated action layer - Business trigger: A named competitor has deployed agent-based routing and reportedly cut on-time misses by ~33%; Northwind leadership is feeling competitive pressure to move from reactive reporting to autonomous/agentic decisioning - Complicating backdrop: SAP ECC-to-S/4HANA migration is already planned for next year, creating a hard constraint around not disrupting that program --- ## 2. Stated Requirements - Real-time / predictive exception handling on inbound freight — alerts must trigger action, not just reporting - Dynamic reslotting within the DCs (automated or AI-driven slot reassignment) - Human-in-the-loop override capability — ops managers must be able to override agent decisions; the system must learn from those overrides ("if my people feel replaced, this dies" — Priya) - Agent-based architecture — Priya used the term "multi-agent" twice unprompted; explicitly wants the system to act, not just report - Data residency compliance for Mexico DCs — order data from Mexican facilities cannot leave Mexico; flagged as a new legal interpretation, not a negotiable position - Pilot live by March (implied: ahead of Q4 next year peak

…output truncated at 250 words.

Step 2 output

# Northwind Foods — RFP Response Proposal Outline Due: October 6 | Pursuit team working document — do not distribute externally --- ## PURSUIT TEAM NOTES BEFORE YOU START Three things to resolve before drafting begins: - Pilot scope must be locked (Open Q3) — Section 4 and 6 cannot be finalized without it - Prior GenAI POC details (Open Q5) — Section 3 and 9 need to know what failed and who ran it - Mexico data residency legal specifics (Open Q1) — Section 3, 7, and 8 all carry unresolved exposure here Assign an owner to each open question before the October 3 internal review. --- ## 1. Executive Summary Extraction references to address: - Competitive urgency — peer firm already running agent-based routing, ~33% on-time improvement - Working capital framing (Marcus / CFO lens) - Human-override adoption risk ("if my people feel replaced, this dies" — Priya) - Pilot live by March; value visible before Q4 peak - Budget signal: $3–5M Y1 What this section must argue or prove: - We understand that the real problem is not bad dashboards — it is a system that watches exceptions happen instead of preventing them; open with that framing, not with our firm's credentials - Lead with the working capital and on-time delivery outcome story, not technology features — this summary will be read by Marcus and must survive a CFO filter - Establish in two sentences that we have a concrete, de-risked path to a live pilot by

…output truncated at 250 words.

Step 3 output

To: priya.ramaswamy@northwindfoods.com From: [Your name] Subject: Northwind / RFP — a few items before we submit October 6 --- Priya, Thanks for the time on September 22 — useful conversation. We're confirmed on submitting a response by October 6. Before we do, I have a small number of questions where your input will materially affect what we propose. I've kept it to six, grouped below. --- Pilot scope and sequencing 1. For the March pilot, are you expecting all three capabilities live — inbound exception handling, dynamic reslotting, and the override-learning loop — or would you prioritize two of three in exchange for a lower-risk, faster go-live? Our current recommendation is to lead with inbound exception handling and the override loop, and phase reslotting into a second wave. Happy to walk through the reasoning if useful. 2. Should the pilot include Mexico DCs, or are you comfortable starting with US locations and expanding once the data residency architecture is validated with your legal team? --- SAP migration 3. What is the current target go-live date for S/4HANA, and which DCs are in the first wave? We want to show Dan explicitly on our timeline where our integration work and his migration program do not touch. --- Mexico data residency 4. Is your legal team's position that order data cannot leave Mexico under any circumstances, or is the constraint specific to certain data types (e.g., personally identifiable information, customer order details)? And is a Mexico-based cloud deployment — compute and storage

…output truncated at 250 words.