A repeatable system for writing email faster without sounding like a robot.
Here is the case for using AI on email and almost nothing else first: email is high-volume, low-stakes-per-message, and formulaic. That combination is exactly where a language model shines.
The volume is real. The average employee now receives 117 emails a day, plus 153 Teams messages, and gets interrupted roughly every two minutes during core hours. [1] Most of those messages are not hard to write. They are just tedious, and they pile up.
Average daily volume per employee: 117 emails and 153 Teams messages. Source: Microsoft Work Trend Index, Breaking Down the Infinite Workday, 2025 [1].
ChatGPT removes the part that actually slows you down, which is not typing. It’s the blank screen. Deciding how to open, how blunt to be, how to soften a no. Once the words exist on the page, editing them takes a fraction of the time. You go from author to editor, and editing is faster.
The honest rule of thumb: use ChatGPT for the emails that make you hesitate. A one-line reply to a colleague does not need AI. The careful client update, the chase that has to stay friendly, the bad-news message, those do.
Most people type “write an email about the project delay” and get back something bland and generic. Then they conclude AI email is rubbish. The problem is the prompt, not the model.
Use four parts every time. We teach this same structure in our 4-part prompt formula, and it works for almost anything, but email is where it pays off fastest.
Here is a full example you can paste and adapt:
“You’re helping me write a client email. I’m the account lead. The client missed two approval deadlines and our launch is now at risk. I need them to approve the final designs by Friday. Keep it friendly but make the deadline impossible to miss. Under 130 words, plain English, no phrases like ‘just circling back’. Bullet points: designs sent May 28, two reminders since, launch booked for June 19.”
That prompt gives the model everything it needs to write something you would actually send. Notice it tells the AI what to avoid, not just what to do. Banning the phrases you hate is one of the quickest quality wins there is.
These are the messages that eat your afternoon. Swap in your details and go.
“Write a short follow-up email. I emailed [name] last week about [topic] and heard nothing. I want a reply without sounding impatient. Give them an easy out and a clear next step. Under 90 words.”
“Help me decline this request kindly. [Paste the request.] I can’t take it on because [reason]. I want to protect the relationship and maybe offer one alternative. Friendly, brief, no over-apologising.”
“Write an email telling my client we will miss the deadline by one week. Be straight about it, take responsibility without grovelling, and lead with the new plan, not the apology. Calm and professional, under 140 words.”
“Draft a cold-ish intro email to [name, role]. We met briefly at [event]. I want 20 minutes to talk about [topic]. Make the value obvious in the first sentence and make saying yes easy. Under 100 words.”
For long, messy threads, paste the whole thing in and use this: “Here is an email thread. Draft a reply that answers every open question, confirms the next step, and stays warm. Flag anything I should not commit to yet.” It is one of the most useful things you can do with ChatGPT for everyday work.
Default ChatGPT email has a tell. It over-explains, it uses “I hope this email finds you well”, and it ends with “Please don’t hesitate to reach out”. Readers clock it instantly. Here is how to kill the chatbot smell.
And read every draft out loud in your head before sending. If a sentence is one you would never actually say to that person, change it. The model gets you 90 percent of the way. The last 10 percent, the part that sounds like you, is still your job.
If you find yourself typing “keep it under 120 words, plain English, no jargon” into every prompt, stop. Put it somewhere permanent.
ChatGPT’s Custom Instructions let you tell it once how you like your writing, and it applies that to every chat. [2] Add something like: “When you help me write emails, write in British English, keep them under 130 words, use a warm but direct tone, and never use corporate filler phrases.” Our custom instructions setup guide walks through the exact steps.
There is also Memory, which lets ChatGPT remember details across conversations, like your role, your team, and the clients you mention often. [3] It is genuinely useful for email because so much of email is repeated context. Turn it on, then correct it when it remembers something wrong.
Quick caveat worth saying out loud: on a personal account, content you share in ChatGPT can be used to train OpenAI’s models unless you turn off “Improve the model for everyone” in your data controls. [4] Team, Enterprise, and API data is not used for training by default. So do not paste confidential client data, salary details, or anything under NDA into a personal account. Strip names and numbers if you are unsure.
Let’s be honest, a lot of AI email is worse than what people would have written themselves. Usually for one of these reasons.
Pick the one email type you put off most. For a lot of people that’s the chase or the polite no. Take the matching prompt above, paste in a real situation from your inbox right now, and send what comes out (after a quick edit).
Then do the setup step. Five minutes in Custom Instructions saves you re-typing the same rules for the next year. That’s the difference between using ChatGPT as a novelty and building it into how you actually work.
Email is the gateway. Once writing email with AI feels normal, the same muscle moves into reports, updates, and proposals. Start where the volume is highest and the risk is lowest. Start with the inbox.
Use four parts: your role and who you are writing to, the key facts only you know, the outcome you want, and the tone and length. For example: “I’m a project lead emailing a vendor who missed a deadline. Facts: order placed June 1, promised June 5, still not shipped. I want a firm but professional chase that gets a ship date today. Under 110 words, no filler.” The specifics are what turn a generic draft into a sendable one.
Often yes, if you send the raw draft. AI email has tells: phrases like “I hope this email finds you well”, over-explaining, and a slightly formal sameness. You can remove the tell by feeding ChatGPT a few of your real emails to match your voice, banning cliche phrases, and cutting the length by a third. After a quick human edit, it reads as you.
It’s safe for drafting and editing, but be careful what you paste in. On a free or personal account, your inputs may be used to improve the model unless you opt out, so do not paste confidential data, client contracts, or anything under NDA. Remove names and numbers if you are unsure, or use a business or enterprise plan where your data is not used for training.
Paste two or three emails you actually wrote and tell it to match how you open, your sentence length, and your level of formality. For a permanent fix, add your preferences to Custom Instructions so every chat follows them automatically. Turning on Memory also helps, because ChatGPT can remember your role and the people you email often.
All three write strong email, so the better question is which one is already in your workflow. If you live in Gmail and Google Docs, Gemini is right there. If you want the most natural-sounding long replies, many writers prefer Claude. ChatGPT is the most flexible all-rounder and the easiest to set up with Custom Instructions and Memory. Pick the one you will actually open every day.
This guide is based on hands-on training with non-technical professionals who write email all day, alongside Microsoft’s 2025 Work Trend Index on communication overload and OpenAI’s official documentation on Custom Instructions, Memory, and data controls. Every statistic is sourced and linked below.
Weekly tips. Real examples. Practical help for busy professionals.
We care about your data, check out our privacy policy.