How to Use AI for Email Writing (Templates + Examples)

You Spend 2+ Hours a Day on Email. AI Can Cut That in Half.

The average professional sends 40 emails per day and spends about 28% of their workweek managing their inbox. That's over 11 hours a week. Most of that time isn't spent on complex communication. It's spent on routine messages: follow-ups, scheduling, status updates, and polite declines.

AI handles these routine emails well. Here's how to set it up so you actually save time, plus templates you can start using today.

The Right Way to Use AI for Email

Don't paste an email and say "reply to this." That gives the AI no context about your relationship with the sender, your priorities, or your tone. Instead, follow this pattern:

Context + Situation + Desired Outcome + Tone

Example: "I'm replying to a client who asked for a project timeline update. We're two weeks behind schedule because of a vendor delay. I want to be transparent about the delay but reassure them we have a plan to catch up. Keep the tone professional but warm."

That's 30 seconds of context that turns a 10-minute drafting exercise into a 1-minute review-and-send.

Template 1: Professional Follow-Up

Prompt: "Write a follow-up email to [name/role] about [topic]. We last spoke [timeframe]. I want to [goal]. Keep it under 100 words and friendly but professional."

This works for sales follow-ups, post-meeting check-ins, and nudging someone who hasn't responded. The key is specifying the length. AI emails tend to be too long by default. Telling it "under 100 words" forces conciseness.

Template 2: Cold Outreach

Prompt: "Write a cold outreach email to [role] at [type of company]. I'm offering [product/service]. The main value prop is [specific benefit]. Include one specific observation about their company/industry to show this isn't a mass email. Keep it under 150 words with a soft CTA."

Cold emails are tricky because they need to feel personal. The "specific observation" instruction forces the AI to generate something that doesn't read like a template, even though it is one. You'll want to customize this observation with real details before sending.

Template 3: Delivering Bad News

Prompt: "Write an email to [recipient] explaining that [bad news]. The reason is [context]. Here's what we're doing about it: [solution/next steps]. Be direct and don't over-apologize. Professional tone."

Most people over-apologize in bad news emails, which makes them longer and more awkward than necessary. The instruction "don't over-apologize" produces emails that acknowledge the issue, explain the path forward, and move on. That's what recipients actually want.

Template 4: Saying No Politely

Prompt: "Write a polite decline email for [request]. I'm saying no because [reason, or just 'I'm not available']. If appropriate, suggest an alternative. Keep it brief and warm."

Saying no is hard for most people, which is why these emails take disproportionate time to write. Having AI draft the decline removes the emotional friction. You review, adjust, and send. The whole thing takes two minutes.

Template 5: Meeting Request

Prompt: "Write a meeting request email to [person/group]. The purpose is [topic]. I'd like to cover [2-3 agenda items]. Suggest [timeframe] and mention that I'm flexible on timing. Include the estimated duration."

Good meeting requests include an agenda and a time estimate. Most people skip both. This template forces both into the email, which makes the recipient more likely to accept because they know what they're committing to.

Template 6: Thank You / Appreciation

Prompt: "Write a thank you email to [person] for [specific thing they did]. Mention the specific impact it had on [project/team/me]. Keep it genuine and under 80 words."

Thank you emails are important but easy to put off. Having a quick way to generate them means you actually send them. The "specific impact" instruction prevents generic flattery and makes the appreciation feel real.

Tips for Better AI Emails

Always review before sending. AI occasionally produces something that sounds slightly off or misses context only you know. A 30-second review catches these issues. Never send an AI email without reading it first.

Build a personal style guide. If you always sign off with "Best," tell the AI. If you never use exclamation points, mention it. Over time, you can build a system prompt that captures your email voice so every draft feels like you.

Use the right model. For email writing, models with strong conversational abilities work best. GPT-5 and Claude both handle email tone well, though they have slightly different styles. Claude tends to be more measured; GPT-5 is more casual. Try both and see which matches your communication style.

Batch your email time. Instead of responding to emails throughout the day, batch them. Open 5-10 emails, generate drafts for all of them, then review and send in one session. This is faster than switching between email and AI for each message.

A Faster Email Workflow

Here's the workflow I use daily:

1. Scan inbox and star emails that need responses
2. Open AI (I use Admix because I can switch models based on the email type)
3. Generate drafts for all starred emails using the templates above
4. Review each draft, make edits, and send
5. Total time: about 20 minutes for 10-15 emails

Without AI, the same batch would take 45-60 minutes. That's a real, measurable time saving every single workday. Over a year, it adds up to weeks of recovered time.