Skip to main content
0

How to Use AI to Write Better Emails (Without Sounding Like a Robot)

A
a-gnt5 min read

AI can help you write emails faster and clearer. But if you're not careful, every message sounds like it was written by a corporate press release. Here's how to avoid that.

The Problem with AI Emails

You've seen them. You've probably received them. Emails that start with "I hope this email finds you well" followed by three paragraphs of perfectly structured, completely soulless prose that ends with "Please don't hesitate to reach out."

That's what happens when someone pastes "write me an email" into an AI and hits send without editing. The AI produces text that is technically correct and emotionally vacant. It reads like a template because it essentially is one — an average of every professional email in the training data.

But here's the thing: AI is genuinely good at email. You just have to use it right.

Step 1: Give It Your Voice

The single biggest mistake people make is not telling the AI how they write. AI defaults to a generic professional tone because that's the safe middle ground. If you want emails that sound like you, you need to show it what "you" sounds like.

Try this approach: paste in 3-4 emails you've actually sent. Tell the AI:

"Here are some emails I've written. Notice my tone, sentence length, and how I open and close messages. Write future emails in this style."

Suddenly your AI-drafted emails will use your contractions, your level of formality, your habits. If you tend to be direct, they'll be direct. If you're warm and chatty, they'll be warm and chatty.

The 📧Email Professional Rewriter prompt does this well. It takes your rough draft and polishes it while keeping your voice intact — it doesn't replace your words with corporate jargon.

Step 2: Draft, Don't Delegate

The best workflow isn't "AI writes the email, I send it." It's "I tell AI what I need to say, AI drafts it, I edit it to sound like me, then I send it."

Think of AI as a first-draft machine. You provide:

  • Who you're writing to and your relationship
  • What you need to communicate
  • Any specific tone requirements (apologetic, firm, casual, formal)
  • Anything you definitely want to include or avoid

The AI generates a draft. You read it, change the parts that don't sound right, add personal details the AI couldn't know, and remove anything that feels generic.

This takes 2 minutes instead of the 15 minutes you'd spend staring at a blank compose window trying to figure out how to word something diplomatically.

Step 3: Use AI for the Hard Emails

Where AI really shines isn't the routine "confirming our meeting on Thursday" emails. It's the ones you dread writing:

The bad news email. "We need to push the launch date back." AI is excellent at finding diplomatic language that delivers bad news clearly without being harsh or evasive.

The follow-up that doesn't sound desperate. "Just circling back" is the most hated phrase in professional email. AI can find genuinely fresh ways to follow up that add value rather than just pinging.

The boundary-setting email. "I can't take on this project." Most people over-explain or over-apologize when setting boundaries. AI can help you be clear, professional, and done in three sentences.

The cold outreach. Whether you're job hunting, pitching a client, or introducing yourself, AI can help craft messages that are personalized and interesting rather than generic and forgettable. The 🎯Elevator Pitch Coach prompt helps with the underlying messaging.

The "I messed up" email. The 🕊️Apology Letter prompt is surprisingly useful here. It helps you acknowledge the mistake, explain what happened, and outline what you're doing about it — without either minimizing or groveling.

Step 4: Trim the Fat

AI emails tend to be too long. Always. The AI is trained to be thorough, which in email translates to saying in four paragraphs what could be said in four sentences.

After you get your AI draft, cut it by 40%. Seriously. Go through and ask yourself: does this sentence add information? If it's just throat-clearing or filler, delete it.

Some AI email crimes to watch for and eliminate:

  • "I wanted to reach out to..." (Just reach out. Skip the narration of your desire to reach out.)
  • "In light of our previous conversation..." (They know about the previous conversation. They were there.)
  • "I believe it would be mutually beneficial if..." (Just say what you're proposing.)
  • "Thank you for your time and consideration." (Fine in a cover letter. Weird in a casual email to a colleague.)

The best professional emails are short, clear, and actionable. AI can get you there — you just need to prune.

Step 5: Match the Medium

An email to your CEO requires different treatment than a message to your friend who happens to be a coworker. AI can adapt, but you have to tell it to.

Be explicit about context:

  • "This is a casual email to a colleague I'm friendly with"
  • "This is a formal email to a client I haven't met"
  • "This is an internal email to my team — keep it brief and direct"
  • "This is a sensitive email about someone's performance"

The tone shift between these is enormous, and AI will nail it if you specify it. Without context, you get the default: a mid-formality email that's appropriate for no specific situation.

Real Examples

Bad AI email:

"Dear Michael, I hope this email finds you well. I am writing to inform you that the project timeline has been adjusted. After careful consideration of the various factors involved, including resource allocation and stakeholder feedback, we have determined that an extension of two weeks would be advisable. I would welcome the opportunity to discuss this further at your earliest convenience. Best regards."

Good AI email (with proper prompting and editing):

"Hey Michael — quick heads up: we're pushing the launch back two weeks. The QA team found some edge cases we don't want to ship with. I've updated the timeline in Notion. Want to grab 15 minutes Thursday to walk through the new plan? — [Your name]"

Same information. One sounds like a legal filing. The other sounds like a human being.

The Tools That Help

A few things on a-gnt that work well for email:

The Golden Rule

If you read your AI-drafted email out loud and it doesn't sound like something you'd actually say, rewrite it until it does. AI is the starting point. You're the editor. The email should sound like you on your most articulate day — not like a robot wearing a suit.

Speed and clarity. That's what AI gives you for email. Use it for the draft. Use your brain for the polish. Your inbox will thank you.

Share this post:

Ratings & Reviews

0.0

out of 5

0 ratings

No reviews yet. Be the first to share your experience.

Related Posts

·5 min read

In the Weeds: Running the Numbers on a Lemonade Stand (With AI Doing the Math)

A dog groomer, a freelance designer, and a bottle of shampoo walk into a spreadsheet. The unsexy, specific, Tuesday-afternoon math that tells a small business owner whether they're making money — and how AI does it in a conversation instead of a cell.

·5 min read

What Your Retired Parents Actually Want From AI (And How to Set It Up in Ten Minutes)

Your mother doesn't need a tutorial on large language models. She needs someone to sit with her and show her how to ask the question she's been carrying around for three weeks. A practical guide for the adult children of curious retirees.

·5 min read

Hacks: The One Prompt That Makes AI Explain Anything Like You're Twelve

Twelve words that turn every AI interaction from a wall of jargon into something you actually understand. Works on taxes, medical results, legal documents, insurance — anything where the gap between expert language and human understanding is the problem.

·5 min read

Hallucinations: The Tool That Rewrites Your Bio and Always Makes It Worse First

A first-person essay by the model, about the specific way my first draft of your bio is always slightly worse than yours — and why that's actually useful information.

·5 min read

Hacks: The Devlog Pacing Trick That Makes a Boring Week Worth Reading

Most devlogs lead with what's next and bury the interesting part. Try the problem-in-the-middle structure instead, and watch a week you were embarrassed about turn into a post people finish.