Automating Email Responses with AI

Email is one of the most time-consuming daily tasks. In this lesson, you will build an automation that reads incoming emails, generates AI-powered draft responses, and delivers them for your review.

What You Will Build

Template 1: AI Email Draft Generator

Trigger: New email arrives in Gmail → AI: Analyze the email and draft a response → Action: Save the draft in Gmail (or send to Slack for review)

This automation handles the thinking and writing. You just review and click send.

Building in Zapier (Step by Step)

Step 1: Set the Trigger

1. Create a new Zap
2. Trigger app: Gmail
3. Trigger event: "New Email"
4. Account: Connect your Gmail account
5. Label/Mailbox: Choose "INBOX" or a specific label like "Needs Reply"
6. Test the trigger to pull in a sample email

Step 2: Add an AI Filter (Optional but Recommended)

Before calling the AI, filter out emails that do not need responses (newsletters, promotions, automated notifications).

1. Add a Filter step
2. Set the condition: From Email does not contain noreply
3. Add another condition: Subject does not contain unsubscribe

This saves AI API costs by only processing emails that actually need a response.

Step 3: Add the AI Action

1. Add OpenAI → "Conversation"
2. Model: gpt-4o-mini
3. System prompt:

You are a professional email assistant. Your job is to draft concise, friendly email responses.

Rules:
- Keep responses under 150 words
- Match the formality level of the incoming email
- If the email asks a question you cannot answer, acknowledge it and say you will follow up
- Do not make up information or commitments
- End with an appropriate closing (Best regards, Thanks, etc.)
- Do not include a subject line, only the body text

4. User message: Map the Gmail Body Plain field
5. Temperature: 0.5
6. Max Tokens: 300
7. Test the action to see the AI-generated response

Step 4: Deliver the Draft

You have two options:

Option A: Save as Gmail Draft

1. Add Gmail → "Create Draft"
2. To: Map the From Email field from the trigger (reply to sender)
3. Subject: Map Subject and add "Re: " at the beginning
4. Body: Map the OpenAI response
5. The draft appears in your Gmail Drafts folder for review

Option B: Send to Slack for Review

1. Add Slack → "Send Channel Message"
2. Channel: Choose a channel like #email-drafts
3. Message: Combine the original email info and the AI draft:

📧 *New email from:* {From Name} ({From Email})
*Subject:* {Subject}

*Original message:*
{Body Plain}

---
*AI Draft Response:*
{OpenAI Response}

Step 5: Turn It On

1. Review all steps in the Zap editor
2. Click "Publish" to activate your Zap
3. New emails will now be processed automatically

Building in Make (Step by Step)

Step 1: Create the Scenario

1. Create a new scenario
2. Add the first module: Gmail → "Watch Emails"
3. Connect your Gmail account
4. Set the Label to "INBOX" or your preferred label
5. Set Maximum number of results to 5 (processes 5 emails per run)

Step 2: Add a Filter

1. Click the line between the Gmail module and the next module
2. Add a filter: From (email address) does not contain noreply
3. Name the filter: "Skip automated emails"

Step 3: Add the AI Module

1. Add OpenAI → "Create a Completion"
2. Method: Create a Chat Completion
3. Model: gpt-4o-mini
4. Messages:
   • System: Use the same system prompt from the Zapier example above
   • User: Map the Text content field from Gmail
5. Temperature: 0.5
6. Max Tokens: 300

Step 4: Save the Draft

1. Add Gmail → "Create a Draft"
2. To: Map From: email address from the Watch Emails module
3. Subject: Map Subject and prepend "Re: "
4. Content: Map Result from the OpenAI module

Step 5: Schedule and Activate

1. Click the clock icon at the bottom of the scenario
2. Set the schedule (e.g., every 15 minutes)
3. Toggle the scenario ON

Customizing the System Prompt

The system prompt determines the quality of your AI drafts. Here are variations for different use cases:

Customer Support

You are a customer support agent for a SaaS company.
Draft helpful, empathetic responses. If the customer reports a bug,
acknowledge the issue and let them know the team is investigating.
Keep responses under 100 words. Use a warm, professional tone.

Sales Follow-Up

You are a sales development representative.
Draft a follow-up response that addresses the prospect's questions,
highlights relevant product benefits, and suggests a meeting time.
Keep it concise (under 120 words) and end with a clear call to action.

Personal Email

You are a personal email assistant. Draft friendly, casual responses.
Match the tone of the sender. Keep responses brief and natural.
If the email is about scheduling, suggest checking the calendar.

Testing and Refining

After your automation runs for a day or two:

1. Review the drafts - Are they appropriate and accurate?
2. Refine the prompt - If responses are too long, add stricter word limits. If too formal, adjust the tone instructions.
3. Add edge cases - If certain email types get bad responses, add instructions for those specific cases.
4. Monitor costs - Check your OpenAI dashboard to see how much the automation costs per day.

Key Takeaways

• The email automation follows a simple pattern: Trigger (new email) → AI (draft response) → Action (save draft)
• Always add filters to skip automated emails and save API costs
• Use gpt-4o-mini with temperature 0.5 for natural but consistent email drafts
• Customize the system prompt for your specific use case (support, sales, personal)
• Review AI drafts for the first few days and refine your prompt based on the results