Most advice about email says to sort it by date, or by sender, or by “importance.” That never works, because the problem is not the order. The problem is that your inbox is three different jobs wearing the same coat.
One job is quick and boring: confirm a time, answer a simple question. One is hard and emotional: say no, ask for money you are owed, apologize, push back. And one is just noise: newsletters, receipts, long threads you were copied on.
AI is good at all three, but in different ways. So instead of sorting by date, sort by feeling. This post gives you three piles, a prompt for each, and the rules that keep you from sending something embarrassing.
A widely cited McKinsey Global Institute analysis, now more than a decade old, estimated that knowledge workers spend about 28% of their workweek on email. The exact number varies by study and job, but anyone with a full inbox knows the feeling.
Pile One: The Quick Replies
These are the messages that take two minutes to read and ten to answer, because you keep wondering how to phrase them.
The method: give the facts, not the wording. Tell the AI what you want to say in plain words, and let it handle the phrasing.
Copy this prompt:
“Write a short, friendly reply. Here is the email I received: [paste]. I want to say: [your point in one sentence]. Keep it under 80 words. Plain language, no corporate jargon.”
Three habits make the replies better:
- Name the tone. “Warm but brief,” “firm but polite,” or “casual, we are friends.”
- Set a length. AI over-writes by default. “Under 80 words” fixes that.
- Save your best. When you get a reply you love, save it as a template and reuse it.
An illustrative example: a freelance photographer in Denver gets a client asking, for the third time, when the photos will be ready. She types, “Say they will be ready Friday, apologize for the delay once, keep it warm.” Ten seconds later she has three clean sentences. She changes one word and sends it.
Pile Two: The Tough Messages
This is where AI helps most, and where mistakes cost most.
When something stings, a late payment, a rude coworker, a landlord who ignores repairs, we tend to write something too angry or too apologetic. AI can be the calm friend who reads it first.
The method: write it ugly first, then convert it.
- Dump everything into the chat. Write what you really want to say, with all the frustration. This is for your eyes only.
- Then type: “Turn this into a calm, professional message that is clear about what I need and by when. Remove anything that sounds angry or accusatory. Keep the key facts.”
- Read the result. Ask: “Would I be comfortable if this were shown to a third person?”
Never send the ugly version. Writing it out is the point, because once the feeling is on the page, the clear message is easier to see.
Prompts for common hard moments:
- Saying no: “Help me decline this request politely without giving a long excuse. Offer [alternative] if it fits.”
- Chasing a late payment: “Write a firm, friendly reminder that invoice [number] is past due. Mention the due date and ask for a payment date.”
- Apologizing: “Write a sincere apology for [what happened]. Take responsibility, say what I will do differently, and do not make excuses.”
Pile Three: The Noise
This is the cleanup pile, and the easiest win.
Long threads. Paste in a thread and ask: “Summarize this in five bullet points. What was decided, what is still open, and what do I need to do?” Some email apps now do this for you. Google’s support page says Gemini in Gmail can summarize threads, draft and refine emails, and search your inbox, and that it requires an eligible Google Workspace or Google AI plan. Availability changes, so check your own app.
Newsletters and promotions. Spend ten minutes unsubscribing from what you never read. If your app has an AI assistant that can organize mail, it can help with labels and archiving. Google says Gemini will not change your emails until you confirm, and you get 60 seconds to undo. Still, review what it plans to do before you approve.
Searching. Instead of hunting through folders, ask in plain language: “When is my flight on Friday?” or “Find the email about the tax form.” This is where built-in AI assistants shine.
For more on how AI is taking over this kind of small daily task, see the one daily task I will never do without AI again.
The Voice Rule: Do Not Sound Like a Robot
Here is the catch. Anthony Graham, a school administrator with years of HR experience, wrote in Education Week that AI-drafted emails often read as polished but generic, and that recipients may see them as a sign the sender did not put in effort. He also warns that language can come out too defensive or too formal for the relationship.
So treat every draft as a first draft.
- Replace one or two phrases with your own words.
- Add one personal detail only you would know.
- Read it aloud. If you would never say it, change it.
- Cut anything that sounds like a greeting card.
For prompts that help writing sound human, try our guide to AI text generator prompts for human-like content.
The Safety Rules (Read These Before Pasting Anything)
1. Keep sensitive details out. Do not paste medical, legal, financial or personnel details into a general chatbot. Use placeholders: “Employee A,” “Client B,” “[amount].” Graham makes the same point for workplace emails.
2. Check every fact. AI can invent dates, names and policy details. Google itself warns that Gemini “may suggest inaccurate or inappropriate information.” For legal or HR matters, verify with an official source or a professional.
3. You are the sender. If it goes out under your name, you own every word. Always read before you send.
4. Watch for AI-powered scams. Phishing emails can now be well written, so a polished message proves nothing. If an email asks for money or passwords, verify through another channel. Our guides to AI scams and deepfakes and digital fraud protection explain how.
A 15-Minute Inbox Reset
Try this once this week.
- Minutes 1 to 3: Sort the top 20 messages into three piles: quick, tough, noise.
- Minutes 4 to 8: Use the quick-reply prompt on the quick pile.
- Minutes 9 to 12: Pick the one tough message you have been avoiding. Write it ugly, convert it, and sleep on it if you want.
- Minutes 13 to 15: Unsubscribe from five things.
Common Problems and Practical Solutions
Problem: The AI reply sounds fake. Solution: Give it a sample of your own writing and say, “Match this style.” Then edit by hand. Our piece on why most people fail with AI prompts explains why clear instructions matter so much.
Problem: I do not trust it with work emails. Solution: Check your company’s AI policy first. Use placeholders for names and details, or use only the AI tool your employer has approved.
Problem: The tone came out too harsh or too soft. Solution: Tell it exactly what to change: “Make it 20% warmer,” or “Remove the apology and be more direct.”
Problem: I spend all day in meetings and then drown in follow-ups. Solution: Use notes to draft the follow-up. See our guide to AI meeting note takers.
Problem: I still feel behind. Solution: Pair email with a simple plan. Our experiments replacing my to-do list with AI and using AI to plan my entire week show how.
The Takeaway
AI will not make email disappear, but it can take the dread out of it. Sort by feeling. Let AI do the phrasing, and keep the judgment for yourself. For more everyday AI guides, see AI in daily life, the wider look at AI-assisted work, and the AI Overview Search about page.
Frequently Asked Questions
Can AI really write my emails for me? It can write good first drafts. You should still read, correct and personalize them before sending.
Is it safe to paste my emails into AI? Be careful. Remove sensitive details, use placeholders, and follow your employer’s AI policy.
Which AI is best for email? The one you will actually use. Built-in assistants in your email app are convenient, and general chatbots are flexible. Features and plans change, so check each tool’s current options.
What should I never use AI for in email? Anything where an invented fact could cause harm, such as legal, medical or HR matters, unless a professional checks it.
