How to make AI sound like you, not a robot, in every reply

The biggest reason business owners avoid AI is not cost, and it is not complexity. It is embarrassment. They are scared that an AI reply will go out sounding stiff and fake, and a customer who has known them for years will read it and think, that is not him. I get it. Your voice is part of why people trust you. So let me show you exactly how to give an AI your voice, in about ten minutes, with tools you can open right now.

Why AI sounds like a robot by default

Out of the box, ChatGPT and Claude write like a polite stranger in a call centre. Lots of "I hope this message finds you well." Lots of "We sincerely apologise for any inconvenience caused." It is grammatically perfect and completely soulless. That is not a flaw you are stuck with. It is just the average of everything the model has read, and the average sounds like a corporate brochure.

The fix is simple once you see it. The AI is not refusing to sound like you. It just has no idea who you are yet. Nobody told it. So your whole job is to show it.

The one move that fixes everything: show it your real replies

Here is the actual move, and it is almost too easy. Open ChatGPT or Claude (the free version of either is fine for this part). Go into your WhatsApp or your email and find five or six replies you have actually sent to customers. Real ones. Copy them. Then paste them in with a line like this:

"Below are real replies I have sent to my customers. Study how I write: my tone, my length, the words I use, how formal or casual I am. From now on, draft replies in exactly this voice."

That is it. That is the trick the expensive consultants do not want to feel simple. You are not programming anything. You are giving the AI examples, and these models are extraordinarily good at copying a pattern from examples. Five good ones are enough. Ten is plenty.

Notice the honest part here. The AI is not secretly reading your WhatsApp. It cannot see your chats on its own. You are the one choosing which messages to show it and pasting them in. That control is the whole point, and it is a feature, not a limitation.

If you do not want to dig up old messages, write the note

Some owners do not want to scroll through old chats, and that is fine. The shortcut is to write the AI a short note on how you talk. Three or four lines:

"I run a family tailoring business. I am warm but brief. I use the customer's first name. I never use corporate words like 'kindly' or 'esteemed'. I always end by telling them the next step. I write in simple English a busy person can read on a phone."

That little paragraph does most of the work. Combine it with even two real examples and the drafts get very close to you, very fast. Save that note somewhere, because you will reuse it every single time.

Make it permanent so you only do this once

You do not want to paste your examples fresh every morning. The free, set-it-once route is ChatGPT. Open Settings, then "Personalization," then "Custom instructions," and paste your voice note there. Now every new chat already knows how you talk.

Claude has a tidier version of this called Projects, where you save the instructions and your example replies together and every chat inside that project starts in your voice. One honest caveat: Projects is a paid feature, so if you are on Claude's free plan, either do the permanent setup in ChatGPT, or just keep your voice note in a notes app and paste it at the top of each Claude chat. It takes five seconds and works just as well.

If you want it to also know your business facts, the same place works. Your prices, your timings, your return policy, your standard answers. Paste those in once and the drafts stop being generic and start being yours.

The rule that keeps you safe: you still press send

Here is the part I will not let you skip. The AI drafts. You approve. Always. Read every reply before it goes out, especially in the first few weeks while you are learning to trust it. You will catch the odd line that is a little off, you will nudge it, and the AI learns from that too. Keeping your hand on the send button is not a lack of automation. It is what protects your name. Over time you will trust it for the routine stuff and stay hands-on for the delicate stuff, which is exactly the right balance.

The principle, and the honest bit

The whole secret to AI sounding human is that you have to show it the human first. A handful of your real replies, or a short honest note on how you talk, turns a robot into something that sounds like you on a good day. None of this is hidden knowledge, and none of it needs me. You can do this today, for free, before lunch.

What I actually sell is the version where this runs quietly in your business without you fiddling: your voice captured properly, your facts wired in, the drafts landing where your team can approve them in one tap, and the whole thing maintained so it stays sounding like you as your business changes. That build is the paid part, for whoever wants it done properly. The knowledge above is yours to keep either way.

• Puro