How to Use AI to Write Emails Faster

I have this thing where blank email drafts make me anxious. It's not that I don't know what I want to say. It's that starting from nothing, in the right tone, with the right level of formality, without sounding weird or aggressive or too casual, that part paralyzes me. Especially for emails that actually matter. Following up on a job application. Emailing a vendor who messed something up. Telling a client their project is delayed. These are the ones where I'd sit there rewriting the opening line six times.

Using AI for emails changed this almost immediately. I'm genuinely faster now, and more importantly, I feel less stressed about the ones that used to trip me up. Here's exactly how I do it.

First: stop asking AI to write the whole email from nothing

This is the trap people fall into. They give ChatGPT (or Claude, or whatever they're using) one sentence of context and expect a perfect email back. Then they get something that's technically correct but sounds nothing like them, uses weird formal phrases they'd never say, and has to be rewritten anyway.

The better move is to give it more than you think it needs. Before I write a prompt, I spend thirty seconds thinking about: who this is going to, what the relationship is, what I'm trying to accomplish, and any specific details that matter. Then I dump all of that in. It sounds like more work, but it's way faster than the edit cycle you'd do otherwise.

Here's what that looks like in practice. Instead of: "Write an email following up on a job application". I'd write something like: "I applied for a content strategist role at a SaaS company three weeks ago and haven't heard back. I had a brief email exchange with the recruiter before applying and she seemed friendly. I want to follow up without being pushy or weird about it. Keep it short, four or five sentences max. My name is Emily and I mentioned in my application that I'm currently freelancing."

That prompt takes me maybe 45 seconds to write. The email it produces is actually usable. That's the trade I'm making, a little more setup for a lot less editing.

Tone is everything, and you have to be specific about it

Email tone is one of those things that's hard to describe in the abstract but really obvious when it's wrong. I've found that vague tone descriptions like "professional but friendly" still produce pretty bland outputs. What works better is describing the relationship.

"This person is a client I've worked with for two years, we have a good rapport, I've made a mistake and need to tell them" is way more useful than "professional but warm." The model can infer a lot of the right tone from the relationship context. It knows what two-year client relationships feel like. It doesn't know what "professional but warm" means to you specifically.

Another thing I've started doing: pasting in one of my own emails as a style example. Just a paragraph or two. I'll say "here's an example of how I naturally write emails, match this voice" and then give the example. This has been more effective than any tonal description I've tried to write. You can't really describe your own voice, but you can show it.

The emails that used to take the longest

There are a few categories where AI has helped me the most, not the routine stuff, but the emotionally complicated ones where I'd get stuck.

Conflict emails. The ones where someone messed up and I need to address it without burning a relationship. I am not naturally good at these. I either go too soft and the problem doesn't actually get communicated, or I go too direct and it reads as aggressive. Now I tell the AI exactly what happened, what outcome I want, and that I need it to be clear without being harsh. It gives me something in the middle that I can usually work with.

Saying no. I get asked for things I can't or don't want to do and I hate the back-and-forth of finding polite language for it. "Write a polite but firm decline to someone asking me to speak at their event for free. I've done this twice before and need to stop saying yes. I don't want to leave the door open for negotiation." That prompt produces something I can actually send.

Awkward follow-ups. Chasing an invoice. Asking for something a second time. Nudging someone who went quiet. These feel socially loaded and I overthink them. Having AI produce a neutral first draft takes a lot of that weight off. I just edit from there.

The editing step still matters

I want to be honest about this: I always edit what comes out. Not always a lot, sometimes it's just reading it through and changing one or two words that don't sound like me. But I do read it every time before I send it.

AI outputs sometimes have this tell where they're just slightly too polished. A little too structured. The sentences are all grammatically perfect and about the same length. Real emails I write have more variation. A short punchy line. A longer one that explains something. Not every sentence is a complete thought.

So my editing pass is mostly about adding a little texture back in. Cutting a phrase that sounds overly formal. Starting a sentence with "And" if that's what flows naturally. Making sure it sounds like a human wrote it, specifically, like I wrote it.

A few specific prompts that always work

Here are some templates I come back to regularly. I'm sharing the actual structure because the structure is what matters.

• "Write a [short/medium length] email to [who] about [situation]. The relationship is [describe]. Tone should be [describe relationship, not adjectives]. Include [any specific info]. Don't [anything to avoid]."
• "Here's a draft email I wrote. Rewrite it to be clearer and more direct without changing the meaning or making it sound formal."
• "Here's an email I received [paste it]. Write a response that [what I want to accomplish]. Keep it under [word count]."
• "I need to say [the thing I want to communicate] to [person/relationship] but in a way that doesn't [specific concern]. Write me something that does that."

That last one is my go-to for anything where I'm worried about how it'll land. Being explicit about what I'm worried about, "doesn't sound passive aggressive" or "doesn't leave them feeling blamed", does a lot of work.

What I've actually noticed since doing this

My email response time has dropped a lot. More importantly, I've stopped dreading my inbox the same way I used to. The emails that used to sit in my drafts folder for two days because I couldn't find the right words. I knock those out in ten minutes now.

I'm also writing better emails because the AI sometimes models phrasing I wouldn't have thought of, and I steal it. I've added phrases to my personal vocabulary from AI-generated drafts. That's a weird benefit I didn't expect.

The tool doesn't make the decision about what to say. I still do that. But it takes the empty-page friction out of getting words on the screen, and for me, that was most of the problem.