Skip to main content

AI Prompts

AI Prompts for Freelancers: The Conversations That Cost Money

AI prompts for the freelance conversations that cost money: scope creep, raising your rate, a client who has gone quiet, and saying no without burning it.

Written bySaad Amjad
Updated
Read
~5 min
A laptop and notebook on a plain desk, set up for a freelancer's working session.

The prompts worth having as a freelancer are not the ones that write your social captions. They are the ones that get you through the four or five conversations you keep putting off — the scope-creep email, the rate rise, the vague brief, the polite no.

Those are where money actually moves. They are also the messages people rewrite six times and still send badly, because the hard part is not the words. It is holding three things in one paragraph without sounding either apologetic or aggressive.

Before you use any of these

Three things, and the third is the one that changes the output most.

  • Paste the real thread. These prompts are written to take the actual email chain or brief. A prompt working from your summary of a conversation produces a reply to your summary.
  • Say what you want to happen. Not “write a reply” — “write a reply that keeps this client and gets the extra work paid for”. The outcome is the instruction.
  • Name what you never say. Your own anti-patterns, listed explicitly. This is what stops the output reading as generic professional email, and it works better than any request for a particular tone.

1. Scope creep on a fixed price

The one that costs the most, because the alternative to sending it is doing the work for free.

The request that is outside what was agreed
Here is the original scope we agreed: [paste scope]. Here is what the client has just asked for: [paste request]. Write a short reply that does three things in this order — confirms I can do it, states plainly that it sits outside the agreed scope, and offers two routes forward: a fixed price for the addition, or swapping it against something currently in scope. Under 150 words. Businesslike and warm, not apologetic, and no suggestion that they did anything wrong. Do not use the word "unfortunately". End with a direct question so the next move is theirs.

Why it works: the ordering is the whole thing. Leading with “yes I can do that” means the scope point arrives as information rather than as a refusal, and offering the swap gives a client with a fixed budget a way to say yes. Banning “unfortunately” removes the single most common tell of a message written reluctantly.

Limitation: it cannot tell you whether the request really is out of scope. If your original scope was vague, no email fixes that, and the honest answer may be that you absorb this one and write a tighter scope next time.

2. Raising your rate with an existing client

A rate increase on a current account
I am raising my rate for an existing client from [old] to [new], effective [date]. We have worked together [length of time] and the work is [describe it briefly]. Write a short message — under 120 words — that gives the new rate and the date it starts, makes clear it applies to work from that date rather than retroactively, and does not justify the increase at length. One short sentence of context at most. Confident, not defensive. No apology, no gratitude padding, and do not invite them to negotiate. Offer a call only if they want to talk it through.

Why it works: the instinct is to explain, and explaining is what weakens it. Every reason you list is a reason they can dispute, so a message with four justifications has given them four openings. The date and the scope of the change are the only two things that must be unambiguous.

Limitation: it will not tell you whether the new number is right, and it has no idea how they will react. If this client is a third of your income, the decision deserves more thought than the email does.

3. The brief that is too vague to price

Questions to ask before quoting
Here is a brief I have been sent: [paste brief]. Do not write a proposal and do not estimate a price. Instead, list the eight questions I must have answered before I can quote this safely, ordered by how much each one could change the price. For each, add one line on what a bad answer would tell me. Then separately list anything in the brief that is ambiguous enough that two reasonable people would read it differently.

Why it works: it uses AI for the part it is actually good at — spotting what is missing — while keeping the pricing with the person who knows the market. The last request is the valuable one: ambiguity in a brief becomes a dispute in week three, and it is far easier to see before you have emotionally committed to the job.

Limitation: the questions are the common ones. The question specific to your trade, the one that comes from having been burned, is still yours to add.

4. A client who has gone quiet after a proposal

Following up on a proposal, once
I sent this proposal [number] days ago and have heard nothing: [paste proposal summary]. Here is our prior relationship: [describe, or say this is a new prospect]. Write a follow-up under 80 words. It should make replying easy rather than make them feel chased, give them a straightforward way to say no, and include one new piece of value or a relevant observation rather than only repeating the ask. No guilt, no false deadline, no "just circling back". This is my last follow-up, so it should close cleanly either way.

Why it works: explicitly offering the no is what gets replies, because most silence is avoidance rather than disinterest. Telling it this is the last follow-up produces a message that can end the thread with dignity instead of leaving it hanging for another month.

Limitation: no email overcomes a proposal that was wrong for them. If two of these have gone quiet in a row, the problem is upstream.

5. Saying no without burning it

Declining work you do not want
A client I value has asked me to take on this: [describe the work]. I do not want it because [real reason]. Write a decline under 100 words that does not use my real reason if it would be unflattering, does not invent a fake scheduling excuse, keeps the door open for the kind of work I do want, and names that kind of work specifically. Warm and brief. Do not offer to reconsider, and do not recommend anyone unless I say so.

Why it works: the two bans are the point. A fake excuse has to be remembered and maintained; a vague no leaves them guessing what you do want. Naming the work you would say yes to turns a decline into positioning, which is the only version of this message that pays you back later.

Limitation: it cannot judge whether you should actually say no. If you are declining because the client is difficult rather than because the work is wrong, that is a bigger decision than this email.

6. Ending an engagement cleanly

Winding down, with a handover
I am ending a [length] engagement with a client. Reason to communicate: [reason, or "moving on from this kind of work"]. Outstanding items: [list]. Write two things. First, a short message giving a final date, naming what I will complete before then, and what they will need to arrange themselves. Second, a handover checklist of everything they should have from me — files, access, credentials, documentation, anything in my tools they will lose access to. Professional and unresentful. No door left half open.

Why it works: the second half is the part people forget and then spend six weeks being emailed about. Asking for the handover list while you still have everything in front of you is worth far more than the message itself.

Limitation: the list comes from what you told it. Walk your own tools afterwards — the access nobody remembers is the one nobody mentioned.

What these cannot do

Each prompt above names its own limitation, and they share three.

They do not know the relationship. A message that is exactly right for a client of four years is too familiar for one of four weeks, and the only person who can tell is you. They cannot price anything, as covered above. And they cannot make the decision the message communicates — whether to hold the line on scope, what the new rate should be, whether this client is worth keeping.

That last one matters. The reason these conversations are hard is rarely that the wording is difficult. It is that you have not decided yet, and a well-drafted email is a comfortable way to avoid noticing. Decide first, then draft.

How to use this page

Best for
Pick the one conversation you are currently avoiding, run that prompt with the real thread pasted in, and send an edited version today. One of these sent is worth more than all six saved
Not for
Collecting prompts. A prompt library you never open is the same as no prompt library, and these only work with your real context in them

Where to go next

Share this guide

XLinkedIn

Questions

Frequently asked questions

  • What are the most useful AI prompts for a freelancer?

    Not the content ones. The prompts that return the most are for the conversations you put off — telling a client that a request is outside the agreed scope, raising your rate on an existing account, asking the questions that a vague brief needs before you price it, and declining work politely. Those are the exchanges where a badly worded email either costs you money or costs you the client, and they are also the ones most people rewrite six times before sending.

  • Why do most freelancer prompt lists not help much?

    Because they optimise for count rather than for the moments that matter. A list of forty prompts for social captions and blog intros is solving a problem you can already solve. The expensive problems are relational — scope, price, saying no — and each needs a prompt that supplies real context about the relationship rather than a template with a blank in it.

  • Should I send what the AI writes?

    Not as it comes. Treat every output here as a draft that has removed the hardest part, which is starting. Read it as the client will read it, cut anything that sounds like a policy rather than a person, and put back the specific detail only you know — the history, the favour they did you in March, the thing you actually want to happen next. The draft saves you the blank page, not the judgement.

  • How do I handle scope creep without annoying the client?

    Separate the three things the message has to do, which is what makes it hard to write cold. Confirm you can do the extra work, state plainly that it sits outside what was agreed, and offer a route forward rather than a complaint — a price for the addition, or a swap against something already in scope. Most scope-creep emails fail because they do only the second of those, which reads as an accusation.

  • What is the best way to tell a client my rates are going up?

    Early, briefly, and without a justification essay. Give a date the new rate starts, make clear it applies from then rather than retroactively, and do not list your reasons at length — a long explanation invites a negotiation about each reason in turn. The prompt on this page is written to resist that instinct, because the instinct is strong and it is what weakens the message.

  • Can AI tell me what to charge?

    No, and treat any number it offers with suspicion. It has no information about your market, your pipeline, how badly you need this particular job, or what this client has paid before. It is genuinely useful for pressure-testing your reasoning once you have picked a number, and for writing the message that communicates it. The number itself is yours.

  • Are there prompts for quoting and chasing invoices on this page?

    No, deliberately. Those have their own home — quotes, pricing and proposals plus the money conversations are covered in our ChatGPT prompts for small business guide, and the automated side of invoice follow-up has a page of its own. This page covers the conversations that are specific to freelance client relationships rather than to selling and billing in general.

  • How do I stop the output sounding like AI wrote it?

    Give it something to imitate and something to avoid. Paste two or three paragraphs of your own real emails and ask it to match that voice, then name the things you never do — no exclamation marks, no "I hope this finds you well", no apologising twice. The second half does more work than the first. Most AI-sounding email is not badly written; it is generically polite in a way you are not.