Say you run a small garden-design business. A stranger emails asking for a full backyard redesign by Friday, for half your usual price. You cannot do it. You also do not want to burn the bridge. So the email sits in your inbox for three days while you feel bad about it.
Saying no is one of the hardest parts of running a small business. It is also one of the best uses of AI, because the hard part is not the facts. It is the tone. A good “no” is short, kind, and clear.
Why a “no” feels so hard to write
Most owners make one of two mistakes. Some write a long apology that explains everything and leaves the door half open, so the customer keeps pushing. Others go the opposite way and send two cold lines that read as rude.
A good decline does three small things. It thanks the person, says no clearly, and, if you can, points them somewhere useful. That is all.
Decide your answer first
AI tools (ChatGPT, Claude, and Gemini are the well-known ones) are good at wording. They cannot decide for you whether the job is worth taking. So settle the answer before you open the chat. Ask yourself:
- Is this a “no, never,” a “no, not now,” or a “yes, but on different terms”?
- Do I know someone I could honestly recommend?
- Is there a smaller version of the request I would say yes to?
Write your answers as two or three rough notes. They do not need to be pretty.
A prompt you can copy
Paste this into your AI tool, with your notes underneath:
I run a small business and need to politely turn down a request. Below is the message I received and my notes on why I am saying no. Write a reply of about 90 words. Be warm and clear. Thank them first, say no plainly in one sentence, give one short honest reason (or none if my notes say so), and end with one helpful next step. Do not apologize more than once. Do not promise anything I have not written in my notes. Use plain language, as if I were talking to them across the counter.
[their message here]
[your notes here]
The line about not promising anything matters. Left alone, AI will sometimes add friendly extras such as “I would love to work together in the future” or “I will get back to you next month.” If you do not mean it, delete it.
Three common situations
A job you cannot take. Say you are the garden designer. Your notes might read: “Booked until March. Not interested in half price. Know a landscaper who does quick jobs.” Ask for a reply that offers the landscaper’s name only if you are happy to give it out.
A discount you cannot give. Here the better answer is often a smaller yes. Add to the prompt: “Offer to remove one part of the job to lower the price, instead of cutting my rate.” The customer gets a choice, and you protect your pricing.
A request that keeps coming back. Some customers ask again after you said no. Paste your first reply into the chat and add: “They have asked again. Write a shorter, firmer reply that does not change my answer and does not sound angry.” Firm and polite can sit in the same message.
Check what comes back
AI drafts need a careful read before you send them. Look for these:
- A hidden “maybe.” Phrases like “perhaps in the future” or “let me see what I can do.” If your answer is no, the draft should say no.
- Invented reasons. AI may add an excuse you never gave it, such as “our schedule is full.” If it is not true, cut it. Honest and short beats clever.
- A referral you cannot stand behind. Only name another business you have actually worked with or trust. Do not let AI suggest one, because it can make up names.
- Too much sorry. One apology is plenty. More than that sounds like you are asking permission.
Keep a few of your own
Once you have a decline you like, save it. Next time, paste it back and say: “Here is a decline I like. Using the same tone and length, write one for this new request.” After a few rounds you will have a small set for the cases you see most, and a no will take two minutes instead of three days.
Try this today
Find the one message you have been avoiding replying to because the answer is no. Write two or three rough notes about your answer, paste them into a free AI chat with the prompt in this guide, and read the reply aloud. Remove anything that is not true or not what you mean, then send it. You will probably feel lighter.