You write between 40 and 80 emails per day. Some are one-line replies. Some are 500-word sales pitches. Some are newsletters that go to thousands of subscribers. All of them take longer than they should, and most of them sound worse than you want them to.
AI email writing is not about removing humans from email. It is about removing the blank-page problem. The hardest part of any email is the first draft -- staring at an empty compose window trying to figure out how to start, what to include, and how to end without sounding either too casual or too corporate. AI handles that mechanical first pass in seconds. Your job becomes editing and adding the human elements that make emails actually work.
This guide covers every email type you send, with specific AI prompt templates for each, the tools worth paying for, and the techniques that keep AI-written emails from sounding like AI-written emails.
The Email Types and How AI Handles Each One
Not all emails benefit equally from AI. Some types are 90 percent automatable. Others need significant human input. Knowing the difference saves you from over-relying on AI where it falls short and under-using it where it excels.
Cold Outreach Emails
This is the hardest email type for AI because cold outreach depends on specificity. A generic cold email gets deleted. A personalized one gets read. AI can handle the structure but not the personalization -- that requires research you need to do yourself.
The AI role: Generate the email framework, suggest different angles and hooks, and handle the body copy after you provide the personalization elements.
The human role: Research the recipient, identify the specific reason for reaching out, and write or customize the opening line.
Prompt template:
Write a cold outreach email to a [job title] at a [company type/size]. Context: [why you are reaching out -- something specific about their company, a trigger event, or a mutual connection]. My product/service: [one-sentence description]. The specific problem I solve for them: [be precise]. Constraints: under 100 words total. No corporate jargon. No "I hope this finds you well." The opening line must reference the specific context I provided. End with a low-commitment CTA: [book a 15-min call / watch a 2-min demo / reply with a yes if interested]. Write 3 variations with different angles.
What to edit: Always rewrite the opening line with genuine specifics. Add one detail that proves you actually looked at their company. Remove any sentence that could apply to any recipient -- if swapping in a different company name does not break the sentence, it is too generic.
Newsletter Emails
Newsletters are AI's sweet spot. The structure is predictable: introduction, content sections, CTAs, sign-off. AI produces strong first drafts because newsletters follow patterns, and patterns are what AI does best.
The AI role: Generate the full newsletter draft from your content inputs, write section transitions, create compelling preview text, and produce subject line options.
The human role: Provide the content topics, add personal commentary and opinions, ensure factual accuracy, and inject your voice into the introduction.
Prompt template:
Write a newsletter email for [audience description]. This week's content: [list 3-5 topics, links, or key points to cover]. Newsletter name: [name]. Tone: [describe -- e.g., "conversational, slightly irreverent, like talking to a smart colleague"]. Structure: brief personal intro (2-3 sentences), then a section for each topic (50-80 words each) with a clear transition between them, then a sign-off. Total length: [300-500 words]. Generate 3 subject line options that are under 50 characters and do not use clickbait tactics.
What to edit: The introduction is where your personality lives. Rewrite it completely with a personal anecdote, observation, or opinion related to the week's theme. Edit transitions between sections to sound less formulaic. Remove any sentence that reads like filler.
Transactional Emails
Order confirmations, shipping updates, password resets, account notifications -- these are high-volume, low-creativity emails that AI handles almost perfectly. The structure is standardized, the information is factual, and the tone should be clear and helpful.
The AI role: Generate templates for each transactional email type that you then populate with dynamic variables from your system.
Prompt template:
Write a [order confirmation / shipping update / password reset / trial expiration] email template. The email should be: clear (the reader knows exactly what happened and what to do next), brief (under 100 words of body copy), and helpful (include any relevant next steps or support information). Tone: warm but efficient. Include placeholder variables in brackets for [customer name], [order number], [relevant details]. Do not be overly friendly -- this is a functional email. No exclamation marks.
What to edit: Minimal. Transactional emails benefit from consistency, so once you have a template you like, use it across all messages of that type. Review quarterly for any language that has become stale.
Customer Support Emails
Support emails require empathy, accuracy, and problem-solving. AI can handle the structure and tone, but you must verify that the solution or information provided is correct. AI confidently generates plausible-sounding answers that may be wrong.
The AI role: Draft the response framework, suggest appropriate tone and language, and structure the solution clearly.
The human role: Verify the accuracy of any troubleshooting steps or information. Add specific details about the customer's situation.
Prompt template:
Write a customer support reply. Customer issue: [describe the problem they reported]. The solution/answer: [provide the actual correct information or steps]. Customer's tone: [frustrated / confused / just asking a question]. Our tone: empathetic but not apologetic to the point of being obsequious. Structure: acknowledge their issue (1 sentence), provide the solution (numbered steps if applicable), offer next steps if the solution does not work. Under 150 words.
What to edit: Always verify the technical accuracy of any steps or information. Check that the tone matches the customer's emotional state -- an overly cheerful response to a frustrated customer makes things worse.
Follow-Up Emails
Follow-ups are uncomfortable to write because they require persistence without pestering. AI removes the emotional friction by generating options quickly, so you are choosing between drafts rather than agonizing over a blank page.
The AI role: Generate follow-up variations at different levels of assertiveness and with different value-add angles.
Prompt template:
Write a follow-up email. Context: I sent [describe original email] to [recipient] [timeframe] ago and received no response. This is follow-up number [1/2/3]. The goal: [get a reply / schedule a call / close a deal]. Write 3 variations: one that adds new value (a resource, insight, or data point relevant to them), one that is brief and direct (under 50 words), and one that gives them an easy out ("If the timing is not right, totally understand -- just let me know and I will not follow up again"). No guilt-tripping. No "just checking in" or "just bumping this to the top of your inbox."
What to edit: Pick the variation that matches the relationship dynamics. If this is a warm lead who has engaged before, the direct version works. If it is cold outreach follow-up, the value-add version gives them a reason to reply.
Subject Line Generation with AI
The subject line determines whether your email gets opened. AI is genuinely good at this because subject line performance is data-driven -- patterns in what gets opened are learnable.
The Generation Process
Do not generate one subject line. Generate twenty. Then filter.
Prompt:
Generate 20 subject line options for an email about [topic] to [audience]. Mix the following approaches: curiosity gap (make them wonder), direct benefit (tell them exactly what they get), personalization (reference their situation), urgency (time-sensitive without being fake), and social proof (reference peers or data). Keep all options under 50 characters. Do not use ALL CAPS, excessive punctuation, or words that trigger spam filters (free, guaranteed, act now).
The Filtering Process
From twenty options, filter down to three using these criteria:
- Would I open this? Not "is this clever" but "would I actually click on this subject line in my inbox at 9 AM on a Tuesday?"
- Is it honest? Does the subject line accurately represent what is inside the email? Misleading subject lines boost open rates and destroy trust.
- Is it specific? "3 things we changed in your dashboard" beats "Exciting updates inside." Specificity signals value.
A/B Testing Subject Lines with AI
Use AI to create structured A/B tests rather than random variations.
The testing framework:
Generate two subject lines that differ on exactly one dimension:
- Length test: Short (under 30 characters) vs. standard (30-50 characters)
- Personalization test: With recipient name/company vs. without
- Approach test: Question vs. statement
- Specificity test: General benefit vs. specific number/result
Prompt:
I want to A/B test subject lines for [email description]. Create 5 paired options where each pair differs on exactly one variable. Label each pair with the variable being tested. Keep all options under 50 characters.
After each test, record the winner and the variable. After 10-15 tests, you will have a clear picture of what your specific audience responds to. Feed these learnings back to AI for increasingly targeted subject line generation.
Maintaining Your Voice: The Voice Training Protocol
The number one complaint about AI-written emails is that they sound like AI-written emails. Here is the protocol that fixes this.
Step 1: Build Your Voice File
Collect 15-20 of your best emails. These should be emails where you received positive responses, closed deals, or got compliments on your communication. Paste them into a document. This is your voice training set.
Step 2: Extract Your Patterns
Feed the voice file to AI with this prompt:
Analyze these emails and identify the author's writing style. Describe: average sentence length, vocabulary level, use of humor or directness, common phrases or expressions, how they open and close emails, their approach to persuasion, and any distinctive patterns. Format this as a style guide I can reference in future prompts.
Step 3: Use the Style Guide
Include the extracted style guide in every email writing prompt. Either paste it directly or (if using ChatGPT) save it in a Custom GPT.
Using the following style guide, write [email description]: [paste style guide]
Step 4: The Three-Sentence Test
After AI generates any email, read the first three sentences aloud. If they sound generic -- if you could swap in anyone's name as the sender without noticing -- rewrite them. The opening of any email is where voice matters most. The middle and end can tolerate more AI-generated structure because the reader has already committed to reading.
The Tools Worth Paying For
ChatGPT / Claude (General-Purpose)
Best for: All email types when you need flexibility and quality.
Both produce excellent email drafts when prompted well. ChatGPT is faster for quick variations. Claude produces more nuanced, on-brand output for longer emails. Use the prompts in this guide directly with either tool.
Cost: $20/month each for the pro versions. Either one is sufficient. Both together give you the widest capability.
Lavender
Best for: Sales outreach specifically.
Lavender sits inside your email client and scores your emails in real time based on data from millions of sales emails. It tells you if your email is too long, if your reading level is too high, if your subject line is weak, and if your CTA is unclear. The AI writing suggestions are focused on sales effectiveness, not general quality.
Cost: $29/month for the individual plan. Worth it if you send more than 20 cold emails per week.
What makes it different: Lavender does not just write emails -- it scores them against benchmarks. Knowing that your email scores a 72 out of 100 and that shortening it by two sentences would raise it to 85 is actionable in a way that general AI writing is not.
Flowrite
Best for: Quick daily emails -- the ones that are not marketing but still take time.
Flowrite generates contextual email replies from brief instructions. You type "decline politely, suggest next quarter" and it produces a full reply. It works inside Gmail and Outlook, so you do not need to switch tools.
Cost: Starting at $5/month. Worth it if you send 30+ routine emails per day.
What makes it different: Speed. Flowrite is not for crafting a perfect sales email. It is for the 30 emails per day that are necessary but not strategic -- scheduling, confirming, declining, updating, acknowledging.
Your ESP's Built-in AI (Klaviyo, Mailchimp, ActiveCampaign)
Best for: Marketing campaign emails.
These tools have the advantage of context -- they know your subscriber data, past email performance, and templates. Their AI features generate content that fits your existing email design and references subscriber behavior. The quality of AI writing varies (Klaviyo is strongest, Mailchimp is weakest), but the integration advantage is significant.
Cost: Included in your ESP subscription.
What makes it different: Integration with your subscriber data means the AI can reference purchase history, engagement patterns, and segmentation. General-purpose AI tools cannot do this without extensive manual context.
The Editing Checklist
Every AI-generated email should pass through this checklist before sending.
The opener test: Does the first sentence sound like you wrote it, or could it have come from anyone? If anyone, rewrite it.
The specificity test: Does the email contain at least one detail that is specific to the recipient, the situation, or your experience? If not, add one.
The length test: Is the email as short as it can be while still containing everything the reader needs? AI tends to over-explain. Cut ruthlessly.
The tone test: Read it aloud. Does it sound like how you talk? If it sounds formal where you are casual, or casual where you are formal, adjust.
The CTA test: Is there exactly one clear action you want the reader to take? AI sometimes buries the CTA or includes multiple competing asks.
The cringe test: Is there any sentence that makes you wince? AI occasionally produces phrases that are technically fine but feel off -- overly enthusiastic, weirdly corporate, or tone-deaf. Trust your instinct and cut or rewrite those sentences.
Building an Email Template Library
The long-term efficiency gain from AI email writing comes from building a template library. Instead of starting from scratch each time, you build and refine templates for your most common email types.
The Process
- Use AI to generate a first draft for each common email type (welcome, follow-up, cold outreach, newsletter, etc.)
- Edit the draft until it matches your voice and standards
- Save it as a template with clear variable markers: [recipient name], [specific detail], [product name], etc.
- Use the template for future emails, customizing only the variable sections
- Every quarter, review template performance and ask AI to suggest improvements based on response data
What This Looks Like in Practice
After building your template library, the workflow for most emails becomes:
- Identify the email type (30 seconds)
- Pull the template (10 seconds)
- Fill in the variables and personalize (2-5 minutes)
- Review and send (1-2 minutes)
Total: 3-8 minutes per email, down from 15-30 minutes of writing from scratch.
For a person sending 40 emails per day, even modest time savings per email add up to hours reclaimed each week. Those hours go to the work that actually matters -- strategy, relationship building, and the creative thinking that no AI can automate.
Start Here
Pick the email type you write most often. Use the corresponding prompt template from this guide. Generate a draft. Edit it using the checklist. Send it. Track the response.
That is one email. Do it ten times and you will have a working AI email workflow that cuts your writing time in half while keeping your voice intact. The tools and templates come later. The habit of using AI as a drafting partner -- not a replacement -- starts with one email today.