ChatGPT prompt for writing a client proposal

This is the prompt we run after every discovery call, before writing the proposal. Its core rule: the AI invents no need the prospect never expressed and leaves pricing blank. It lists what it does not know instead of padding the gaps.

The prompt to copy

From these discovery notes, fill in a proposal skeleton for {prospect}:
1. Their situation & goal (their words where possible)
2. The problem, framed by business impact
3. Proposed approach — phases with what/why (match against our services: {list})
4. What we need from them
5. Timeline assumption
6. Three questions I should resolve before sending.
Leave pricing blank. Do not invent needs they didn't express. Notes: {paste}

Replace the fields in curly braces (the prospect’s name, your list of services), then paste your notes exactly as they are: typos, fragments, and shorthand included.

When to use it

After a sales or discovery call, before writing the proposal, ideally the same day while the notes are fresh. It is the first link in a short chain: this prompt produces the skeleton, a second prompt turns it into prose once your questions are resolved. Three hours of dreaded writing become thirty minutes of review.

What you get back

1. The prospect’s situation and goal, in their words.
2. The problem framed by business impact, not jargon.
3. A phased approach, matched against your actual services.
4. The list of what you need from them.
5. A timeline assumption.
6. Three questions to resolve before sending, the ones that usually sink a proposal.

Real example: discovery notes → proposal skeleton

Input: Maya’s notes from a discovery call with Marlow & Daughters, a regional bakery chain (11 locations) that wants a rebrand. An excerpt, as-is:

problem in ruth's words: "we look like a village bakery but we're becoming a regional brand and it shows — every shop sign is slightly different"
dev mentioned website is "embarrassing" but ruth changed subject — budget thing?
no numbers discussed. dev asked about "typical range", maya deflected to proposal

Output, the part that matters:

6. Resolve before sending:

1. Is the website in or out of scope? Ruth changed the subject when Dev raised it, clarify before proposing, don’t guess.
2. What’s the actual January date? “Fit-out starts January” could mean Jan 2 or Jan 30, the whole timeline hangs on it.
3. Does the board need to see anything, or is Ruth’s sign-off contractually sufficient?

Pricing: (left blank as instructed)

Notice what the AI did not do: it left the website out of scope (Ruth changed the subject, so it became a question, not a line item) and it quoted no price. That is the prompt working, not failing. The human checkpoint, answering the three questions and setting the price, stays in your hands.

The full worked example (complete notes, the entire skeleton, and the prose generated right after) ships inside the kit, with playbook P3 chaining both prompts.