Write a Client Proposal With AI: Structure, Pricing, and Your Voice
Turn a messy brief into a proposal clients actually read: capture inputs once, let ChatGPT draft sections you edit, price honestly in Google Docs, and send from Gmail without sounding like a template mill.
By DoAIWork · October 7, 2026

You do not need a 20-page deck to win a small project. You need a clear proposal that shows you understood the brief, named the deliverables, explained how you work, and put a price on the table without hiding behind vague “custom quote” language.
This guide is for freelancers and small agencies who already have notes from a call, an email thread, or a form submission—and want AI to do the boring assembly, not replace your judgment on scope and pricing.
If you are still turning discovery notes into a scope document, start with Turn discovery call notes into a scope doc clients approve. A proposal is the commercial wrapper around that scope: timeline, terms, and how to say yes.
What you need before you open ChatGPT
Gather these in one place (a Google Doc titled ClientName — proposal inputs works fine):
- Problem in their words — one paragraph copied from their email or call notes.
- Deliverables — bullet list of what they will receive (files, sessions, integrations).
- Constraints — budget hints, deadline, tools they already use, stakeholders who must approve.
- Your boundaries — what is out of scope, revision rounds, support after launch.
- Pricing logic — your day rate or package price, even if you adjust it later.
Without inputs, AI will invent a generic “we will deliver excellence” proposal. With inputs, it becomes a first draft you can defend.
The workflow (human in the loop)
Every step below ends with you editing. The goal is speed with accountability—not autopilot.
Step 1 — Force a proposal outline (not a wall of text)
In ChatGPT, paste your inputs and use a prompt like this:
You are helping a freelancer write a client proposal. Use ONLY the facts below. Output an outline with these headings: Executive summary, Understanding of the problem, Proposed approach, Deliverables, Timeline, Pricing options, Assumptions, Next steps. Flag anything missing as [ASK CLIENT]. Facts: …
Review the outline for hallucinated features. Delete anything you cannot deliver. If [ASK CLIENT] appears more than twice, send a short clarification email before drafting the full doc—cheaper than rewriting later.
Step 2 — Draft sections one at a time
Ask for one section per message, starting with “Understanding of the problem.” Instruct the model:
- Max 120 words per section unless Deliverables or Pricing.
- No buzzwords (“synergy”, “cutting-edge”).
- Mirror the client’s vocabulary from the brief.
- Use “I” or “we” consistently with how you actually work solo vs with subcontractors.
Paste each section into Google Docs under the matching heading. Keep styles simple: Heading 2 for sections, normal text for body, a table for pricing.

Step 3 — Pricing that does not scare you (or the client)
AI is helpful for wording, not for choosing your rate. You pick the number; the model helps you present it.
Use a three-row table in Google Docs:
| Option | What they get | Price | Notes |
|---|---|---|---|
| Core | Minimum viable scope | $X | Payment 50% upfront |
| Recommended | What you actually want to sell | $Y | Includes one revision round |
| Stretch | Nice-to-haves | $Z | Optional phase 2 |
Ask ChatGPT: “Rewrite the Recommended row description in plain English for a non-technical buyer, under 60 words.” You verify math, taxes, and currency.
Step 4 — Terms freelancers forget
Add a short “Assumptions & terms” block (your lawyer beats AI here, but a checklist helps):
- Start date depends on deposit and asset access.
- Client provides feedback within X business days.
- Out-of-scope work is billed at your hourly rate.
- You retain portfolio rights unless NDA says otherwise.
Have ChatGPT tighten the wording; do not let it invent legal guarantees.
Step 5 — Quality pass: the read-aloud test
Before export:
- Read the executive summary out loud. If you stumble, shorten sentences.
- Search for “ensure”, “leverage”, “robust”—replace with verbs you would say on a call.
- Confirm every deliverable has a noun (file, call, dashboard), not a vibe.
- Check that pricing matches the scope doc if you sent one earlier.
Export PDF from Google Docs or share a view-only link if you want live edits.
Step 6 — Send from Gmail without looking automated
Email template:
Subject: Proposal — [Project name]
Hi [Name],
Thanks again for walking me through [problem]. Attached (or linked) is a short proposal with scope, timeline, and pricing options. The Recommended option is what I’d suggest for [outcome].
If anything feels off, reply with one line and I’ll adjust—no need for a call unless you prefer.
[Your name]
Do not paste the entire proposal into the email body; the attachment is the record.
When AI makes proposals worse
Skip full autopilot if:
- The project is regulated (health, finance, legal) and needs professional review.
- You have never done this deliverable before—AI will sound confident anyway.
- The client asked for a fixed RFP format; follow their structure manually.
In those cases, use AI only for outline + clarity, not pricing or compliance claims.
Measuring success
Track three numbers for the next month:
- Time from “call ended” to “proposal sent” (aim under 90 minutes with this flow).
- Win rate on Recommended option vs discounting.
- How often clients ask “what’s included?” after sending (if high, deliverables section was vague).
More practical playbooks live in our guides category.
Quick recap
Capture facts once, outline in ChatGPT, draft section-by-section, assemble and price in Google Docs, sanity-check aloud, send a short Gmail note with PDF or link. AI saves typing; you own scope, price, and tone—and that is exactly what clients pay for.
Related guides

How to Turn Messy Weekly Notes Into a Client Status Email (With AI)
A 15-minute Friday habit: log bullets in Google Docs, draft a clear status email with ChatGPT, edit the facts, and send from Gmail—optional Make reminder included.
Read guide
AI Product Photos Without a Studio: Phone to Marketplace
Phone-to-listing product photos with AI: remove backgrounds, Amazon RGB 255 white mains, Photoroom vs Canva, and mistakes that mislead buyers.
Read guide
Turn discovery call notes into a scope doc clients approve
Same-day one-page scope from your call notes: paste-ready AI prompt, out-of-scope section, and testable deliverables—you send it, not the bot.
Read guide