AI & Future
AI can draft, polish, and shorten emails in seconds. Here is how to use it to write faster while keeping your voice and protecting private details.
The fastest way to write an email with AI is to hand the tool your raw bullet points plus one line specifying tone and length, then edit the draft instead of sending it verbatim. That loop — dump, direct, edit — beats any clever prompt template, because you stay the author of the substance while the model handles the phrasing you were going to labor over anyway.
Most people write a vague prompt ("write an email asking for a deadline extension") and then spend five minutes fixing a draft that guessed wrong. A good prompt has three parts: the facts, the tone, and the length. Give all three at once and the first output is usually 80% done.
A template that consistently works:
Draft an email. Facts: [bullet points, however messy]. Tone: [warm/firm/neutral/apologetic]. Length: [under 4 sentences / one short paragraph]. Do not invent any details I did not give you.
That last clause matters more than it looks. Language models will confidently fill gaps — inventing a specific "2 p.m. Thursday" or a recipient's name you never supplied. Telling it not to fabricate, then scanning the draft for any date, figure, link, or name you did not personally provide, catches the single most common failure of AI-written email.
Input to the tool: "need to push our call from Tuesday to Thursday, apologize for late notice, keep the same 30 min, tone friendly-professional, three sentences." You get back a clean, sendable reschedule note in about two seconds. You then change one phrase so it sounds like you, confirm Thursday is actually free, and hit send. Total time: under a minute, versus the three or four you would have spent staring at the blank reply box.
You do not need a separate app. The biggest time savings come from AI built into the inbox you already use, because there is no copy-paste round trip.
ChatGPT, Claude, and Gemini in a browser tab give you more control over tone and back-and-forth than the built-in buttons, which is worth the copy-paste for anything sensitive or high-stakes. Email-first clients like Superhuman and Shortwave go further, drafting replies from the thread's context and matching your past writing style, but they read your mailbox to do it — a trade-off worth making consciously, not by default.
The same three facts can become a dozen different emails. Vague instructions ("make it nice") get you the model's default register, which trends toward over-polite padding. Specific instructions reshape the whole message.
If the first draft misses, refine — do not restart. "Too formal," "drop the second paragraph," "make the ask clearer up front" each steer it in one exchange, which is faster than rewriting yourself.
Email is where sensitive detail clusters: client names, salary figures, health situations, unannounced plans. Pasting all of it into a consumer AI tool can mean it is retained or used to improve the service. Two habits keep you safe without slowing you down.
In ChatGPT, go to Settings > Data Controls and turn off "Improve the model for everyone," or use a Temporary Chat for one-off drafts. In Claude and Gemini, look for the equivalent training or activity toggle in settings — the exact label shifts as policies update, so check it rather than assume. Enterprise, Team, and API tiers of the major providers generally do not train on your content by default; free consumer tiers are where you need to check.
You can get the same draft while withholding the crown jewels. Ask for "a polite note declining a vendor proposal on price" without naming the vendor or the number, then paste those in yourself afterward. Redact before you prompt: swap real names for "the client," real figures for "[X]." The model writes structure just as well against placeholders, and nothing confidential ever leaves your keyboard.
Recipients who know you notice when an email suddenly does not sound like you. A few habits fix it.
The dead giveaways to delete on sight: "I hope this email finds you well," "I wanted to reach out," "Please don't hesitate to reach out," and stacked connectors like "Furthermore… Moreover… Additionally." Cut the throat-clearing opening entirely and start with the actual point.
The bigger mistake is treating AI as a ghost-writer for messages that must be real. A condolence note, a genuine apology, or a heartfelt thank-you lands precisely because it is yours. Use AI to check spelling and soften a harsh line in those, never to manufacture warmth you do not feel. And the most valuable trick of all: when you have written a furious reply, do not send it — paste it in and ask the model to make it "firm and professional, no anger." That pause has saved more working relationships than any productivity feature.
Only if you send an unedited draft. The tells are structural — generic openers, over-formal register, and a smoothness that does not match your usual voice. Spending fifteen seconds to cut the filler opening and swap one or two phrases into your own words makes the email indistinguishable from something you wrote by hand.
Not by default on a personal free account, because content may be used to improve the model. Turn off "Improve the model for everyone" in Settings > Data Controls, use a Temporary Chat, or better, redact names and numbers first. If your employer provides Microsoft 365 Copilot or a Workspace/enterprise AI plan, those tiers generally do not train on your data, which makes them the safer place for anything confidential.
Use the AI built into your inbox so there is no copy-paste — "Help me write" in Gmail or "Draft with Copilot" in Outlook — and lean on refine chips like Shorten and Formalize for routine replies. Save a standalone chatbot for the handful of high-stakes or delicate messages that deserve a slower, more careful pass.
It can, and you almost never should let it. Auto-reply features draft from thread context, but they cannot know facts only you hold, and they will occasionally invent a commitment you never made. Treat every AI draft as a first pass to approve and edit, not a decision to delegate — the two-second review is what keeps the mistakes out of your sent folder.
Priya translates the fast-moving world of AI and the internet into things you can actually use and understand. She's curious but skeptical, quick to separate genuine progress from hype, and keen to help readers use new tools wisely rather than fearfully.
Keep reading
AI news moves fast and most of it is noise. Here is a calm, jargon-free system for staying informed about what matters without burning out on every headline.
Priya Nadar
Recommendation systems shape what you watch, buy, and read every day. Here is a clear, jargon-free look at how they work and how to stay in control.
Priya Nadar