Make Your AI Drafts Sound Like You

Tuesday, 10:04 p.m. The kids are finally asleep and you’re back at the laptop, rewriting an email AI drafted for you at 4 p.m. It’s polite. It’s correct. It says “I hope this finds you well” and “delve,” and it sounds like every other email in your client’s inbox. You delete half, retype the rest in your own words, and wonder why you used AI at all. The draft took ten seconds. The rewrite took your evening.

AI adoption doesn’t stall because your team can’t get a chat assistant to write. It stalls because what it writes doesn’t sound like the person who has to hit send, so every draft turns into a rewrite. For SMEs building generative AI for business workflows, a ten-minute voice card is an easy first rep: paste three messages you actually sent, ask AI to write your voice in five rules, save it, and start every draft with it. Drafts sound like you. Your evenings stay yours.

This is Tip 27, a life-first framing of the “make AI sound like me” habit people are talking about on X right now. It’s distinct from Tip 16 (voice dump → evening free, which is about speaking your to-dos out loud), Tip 17 (blank reply that doesn’t send itself, which is about getting a first draft started), and Tip 13 (follow-ups that eat evenings). Tip 27 isn’t about what to write or when. It’s about how every draft sounds: three real messages → five voice rules → one saved card → every draft starts with it.

Generic AI drafts don’t save time. They move the work to 10 p.m.

A default AI draft is written for nobody in particular. It reaches for the same openers, the same three-part lists, the same “I’d be happy to help,” because that’s the average of everything it has read. Your clients don’t know you as average. They know your short sentences, your “Talk soon,” the way you get to the point in line one.

Here’s what actually happens. You ask for a draft, skim it, and feel that small wince: not wrong, just not you. You park it “to fix later.” Later is after dinner. Multiply that by five emails, two proposals, and a listing description, and AI hasn’t saved you an hour. It has moved the hour into your evening.

Tip 27 isn’t a new writing tool. It’s a ten-minute setup, using the AI assistant you already have, and you stay the person who reads and sends every message.

Why the “sounds like AI” posts matter this week

People have started noticing AI-sounding phrasing everywhere, including in each other. A trending X story summarizing a Wall Street Journal report describes professionals catching chatbot-style “prompt speak” slipping into meetings, texts, and chats. One recruiter now edits his Slack messages to sound more human, and some offices track AI-like phrases on whiteboards like swear words (X trending: AI chatbot phrases creep into real work conversations). That’s X’s summary of the reporting, and the takeaway for a small business is simple: people can hear generic AI writing, and they notice it.

Practitioners are saying the trust cost out loud. One poster argues that “AI slop is killing trust” and describes setting his AI up to sound like him using a voice style guide built from his own emails and Slack messages, with separate rules per medium and a banned-word list (@maheshboda006). That’s his claim and his setup; Tip 27 borrows only the core idea that your own sent messages are the best training material. A real-estate-focused account that sells voice-training templates says listings “used to sound like everyone else’s” until it started training AI on how it actually writes (@automelle). That’s their claim about their own product; you don’t need a template pack to do the free version below.

You don’t need a fine-tuned model or a new subscription. You need five short rules that describe how you already write, and the habit of putting them in front of every draft.

The three-step habit (paste · write the rules · save and reuse)

Here’s the version I teach in corporate AI training. It works tonight with ChatGPT, Claude, Gemini, Copilot, or whatever assistant your team already uses.

1. Paste three messages you actually sent (redact first)

Pick three different kinds of writing: one client or customer email, one quick text or chat message, and one longer note (an update, a proposal paragraph, a listing description). Choose messages with no client names, account numbers, prices, or confidential details, or redact them first: swap names for [Client], numbers for [amount], and remove anything covered by an NDA, HR, legal, or medical context. If your company has an approved AI tool or a rule about pasting work messages, use that tool and follow the rule.

2. Ask: write my voice in five rules

Don’t ask it to “make it sound like me.” Give it a format:

“Here are three messages I wrote and sent. Names and private details are removed. Write my voice in 5 rules: (1) tone, (2) typical sentence length, (3) words and phrases I actually use, (4) words and phrases I never use, (5) how I open and sign off. Quote a short example from my messages for each rule. Do not invent stories, opinions, or facts about me. Keep the whole card under 120 words.”

The “never use” rule matters most. Banning a handful of filler words fixes more drafts than any clever prompt.

3. Save the card, then start every draft with it

Read the five rules and fix anything that’s off. You know your voice; the AI only guessed. Then save it where you’ll actually use it: a custom instruction or project setting in your assistant, a pinned note, or a text snippet you paste at the top of every request (“Use my voice card: …”). From now on, every draft starts with the card. AI drafts; you read; you send.

Voice rule	What it captures	Example of a rule
Tone	How warm or direct you are	Friendly, direct, no hype
Sentence length	Your rhythm	Mostly short; one idea per sentence
Words I use	Your real vocabulary	“Quick one,” “here’s the plan”
Words I never use	Your banned list	Never “delve,” “leverage,” “I hope this finds you well”
Open and sign-off	How you start and end	First name, straight to the point; “Talk soon, N.”

Generic AI draft	Voice-card draft
Sounds like every other inbox	Sounds like the person who sends it
Rewritten after dinner	Light edit before you hit send
Filler openers and stock phrases	Banned list keeps them out
Different tone every time	Same five rules on every draft
Clients notice it’s AI	Clients just hear you

Team version (a shared voice card, plus everyone’s five rules)

If you run a team, do it in two layers. First, write a company voice card together: five rules for how the business sounds (tone, words you always use for your services, words you never use, how you sign off as a company). Then each person writes their own five rules on top, so a proposal still sounds like the business and an email still sounds like the person who sent it. Review both cards once a quarter and add any phrase customers have flagged as robotic.

This makes a great first exercise for an AI training session or an adoption kickoff. It takes ten minutes, everyone leaves with something they’ll use the next morning, and it teaches the two habits that matter most: give AI good input, and keep a human on every send.

Why SMEs win when drafts sound like a person

In a small business, the voice is the brand. Clients chose you, not a template. Every generic AI email chips away at the thing that makes you different, and every rewrite eats the time AI was supposed to save. Tip 27 doesn’t need a copywriter or a brand agency. It needs three real messages, five rules, and the discipline to start every draft with them.

A card with no review step still sends a robot. A review with no card means rewriting every draft from scratch. Put them together and AI adoption becomes something you can feel by Friday: fewer rewrites, faster sends, and evenings that stay yours. Two guardrails keep it honest. AI drafts and humans send, every time. And never let the AI invent stories, opinions, or experiences “in your voice.” A voice card sets the style, not the facts.

A 30-day voice-card sprint

Week 1: One card, one channel. Build your five rules and use them on every email draft this week. Count how many drafts you sent with only a light edit.

Week 2: Split by medium. Your email voice isn’t your chat voice. Add a short line for texts or Slack, and a line for longer documents.

Week 3: Grow the banned list. Every time you delete a phrase from a draft, add it to “words I never use.” Rerun one old draft and compare.

Week 4: Teach it once. Use corporate AI training to run the team version: one company card, everyone’s five rules, and the redaction rule for picking sample messages. Set a quarterly review; drop anything that felt forced.

Soft next step

If your team has AI tools but still rewrites every draft at night, and clients are starting to hear the template, the gap is usually the habit (good input, a clear voice card, a human on every send), not another writing app. I’m Nathan Graham, founder of Synthetic Echo, Toronto-based · serving Canada & North America. Book a free discovery call and we’ll work out which writing in your business is the best first place for an AI habit, what’s safe to paste, and how training can make AI adoption stick. It’s a prioritization conversation, not a free full audit.