A client sends a short brief, a few messages, and a deadline. You know how to do the work, but turning scattered details into a proposal takes another hour. AI can help organize a first draft if you give it a clean scope and keep the final commitments in your hands.
Start with a short project sheet that you are allowed to share. Include the client’s goal, deliverables, constraints, and questions that still need an answer. Leave out names, account details, unreleased plans, and other confidential material unless the client agreement and the AI service’s approved data terms allow it.
Convert the brief into a scope before drafting
Ask the tool to separate what the client explicitly requested from assumptions and missing details. A proposal built before that distinction is clear can make a guess look like a promise. The National Institute of Standards and Technology’s Generative AI Profile describes how models can present erroneous information confidently, so require the draft to flag uncertainty rather than fill it in.
From this project summary, list:
1. The client's stated goal
2. Requested deliverables
3. Constraints and dates explicitly provided
4. Assumptions that need confirmation
5. Questions I should ask before quoting
Do not invent client facts, results, fees, timelines, or deliverables.
Project summary: [approved, anonymized details]Review the questions before you ask for proposal copy. Clarify anything that changes the work: number of revisions, who supplies materials, what counts as approval, dependencies, and the date the client needs the result. If the client brief is too vague to price, send a short clarification note or offer a paid discovery step.
Draft only the parts the tool can help with
Once scope is clear, ask for an outline with the client’s problem, your approach, deliverables, exclusions, schedule, fee, and next step. Supply your real rates and availability yourself. Ask the model to keep any supplied terms intact and mark missing details with a visible placeholder. Do not let it invent a case study, testimonial, past result, or credential to make the pitch sound stronger.
You could prompt it like this:
Draft a concise proposal using only the project facts and terms below.
Keep the deliverables, exclusions, price, and dates exactly as written.
Use placeholders for anything missing. Do not promise an outcome or add
experience, testimonials, or examples I have not supplied. End with one
clear next step for the client.
Project facts: [approved summary]
My approach: [your method]
Deliverables and exclusions: [your scope]
Fee and schedule: [your terms]Edit the voice until it sounds like you. Then compare each promise with your notes and contract: Is the number of revisions clear? Are dependencies named? Does the fee cover the stated scope? Does the timeline begin when the client provides what you need? The proposal should explain the work you intend to do, not guarantee an outcome you cannot control.
Protect client information deliberately
AI account settings and data terms vary. OpenAI states that business products and its API do not use customer inputs and outputs for model training by default, while personal ChatGPT settings have separate controls. Those terms describe one provider; check the exact service and account you plan to use, and follow your client contract and permission requirements.
Anonymizing a brief also helps you reason about whether the tool needs the original detail. Replace a client name with “retail client”; replace a customer list with a count or a category if the exact data is irrelevant. If you cannot safely share even a reduced version, use AI only for a generic proposal structure and fill in the private facts yourself.
For another fact-first writing workflow, read how job seekers can tailor a resume with AI. If your proposal process involves a kickoff meeting, turn its notes into confirmed follow-ups. After you send the proposal, a short task handoff note can make the next client-work session easier to restart.