Cozyphere

Blog · Published 2026-10-30

How to write a freelance proposal with AI

AI can draft a proposal in seconds. The trouble is that a fast, generic proposal is easy to spot and easy to ignore. Used well, though, an AI assistant can save you the slowest part, turning messy call notes into a clear structure, while you keep control of the parts that win or lose the work: the understanding, the scope and the price. Here is a practical process.

What AI is good at here, and what it isn't

Good at: organising your notes into sections, writing clear plain-English summaries, suggesting phases for a project, tightening wordy paragraphs, and producing several versions of a sentence so you can pick one.

Not good at: knowing what the client actually said, knowing your prices and availability, judging what you can realistically deliver, or knowing which of your past projects are relevant. Left alone, it fills those gaps with confident guesses: invented timelines, generic benefits, sometimes made-up results. That's the part you must own.

So the rule for the whole process is simple: you supply the facts, AI supplies the structure and first draft, you make the final decisions.

Step 1: Get real input before you write anything

A good proposal is mostly a good summary of the client's situation. That comes from a discovery call or a detailed brief, not from AI. Before drafting, make sure your notes answer:

If you can't answer these, go back and ask. No prompt can fix missing information.

Step 2: Give the AI your context once

Write a short paragraph about your business that you can paste at the top of any chat: what you do, for whom, how you like to sound, and what you never promise. Something like:

I'm a freelance [service] working with [type of client]. Typical projects are [scope] over [length]. I write in a plain, warm, direct style. I never promise specific results, rankings or revenue. Proof I can cite: [only real, checkable examples].

Add a rules block for client-facing writing:

Rules: Use only the facts I provide. If something is missing, write [NEED: ...] instead of guessing. Never invent statistics, testimonials, client names, deadlines or prices. Do not promise results.

These two pieces change the quality of the first draft more than any clever wording.

Step 3: Draft from your notes with a clear structure

Clients read proposals in a predictable order: do they understand us, what will they do, when, and for how much. A structure that follows that order:

  1. Situation: their goal and problem, in their words.
  2. Goal: what success looks like for this project.
  3. Approach: three to five phases, each with a deliverable.
  4. Timeline: only dates you can commit to.
  5. Investment: your price, and what it includes.
  6. Not included: the things they might assume are in scope but aren't.
  7. Next step: one clear action, such as approving by a date.

Here's a prompt that puts it together:

Draft a proposal for [CLIENT NAME] from my notes.

My notes: [PASTE DISCOVERY NOTES]
My prices/packages (if any): [PRICES OR 'to be set']

Sections: Situation (their words), Goal, Approach in 3-5 phases with a deliverable each, Timeline (only dates I give you), Investment, What's not included, Next step. Under 600 words, plain language, no hype. Where I haven't given a price or date, write [NEED: price] or [NEED: date]. Do not invent testimonials or results.

The [NEED: …] markers are the important part. They show you exactly where the AI lacked facts, instead of hiding the gaps behind plausible text.

Step 4: Set the price yourself

Don't ask an AI what to charge. It doesn't know your costs, your market or your workload, and "typical rates" it produces may not reflect anything real. Instead, use it as a calculator on your own numbers:

Ask it to show the arithmetic, then check it yourself. If you want to offer options, three packages (a lean version, the recommended version and an extended one) give the client a choice that isn't just yes or no. Make sure you could deliver the biggest one at its price.

Step 5: Write the exclusions carefully

The "not included" section prevents more arguments than any other part of a proposal. Ask the AI to list things a client might reasonably expect but that aren't in your scope, based on the project description. Then edit the list: keep what's relevant, remove what's silly, and add anything from your own experience. Common candidates are extra revision rounds, copywriting or images, ongoing maintenance, training, and work on other platforms.

If the proposal becomes the contract, or sits alongside one, have contract wording reviewed by a professional.

Step 6: Review like the client will

Before you send, read the draft with these questions:

Shorter is usually better. If a section doesn't help the client decide, remove it.

Step 7: Follow up, and handle "too expensive"

Send a short follow-up a few days after the proposal if you haven't heard back. Make it easy to answer: ask whether they have questions, or whether a smaller first phase would help.

If the reply is "that's too expensive", resist the reflex to discount. Three calmer options:

  1. Ask what budget they had in mind and what they're comparing against.
  2. Offer a reduced scope at a lower price, so the price still matches the work.
  3. Hold the price and restate, concretely, what's included.

An AI assistant can draft all three in a minute from their message and your notes on what's flexible. You pick the one you'd be happy to be held to.

Common mistakes with AI proposals

A simple workflow to reuse

Discovery notes → context paragraph and rules → structured draft prompt → your prices and exclusions → review checklist → send → follow up. Once it's set up, a draft takes minutes, and your time goes into the parts that need your judgement.