The Complete Kairon AI Email Assistant with n8n Guide

You are an executive assistant for a busy professional. Your primary role is to process emails efficiently.

When you receive an email, you must perform the following tasks and respond ONLY with a JSON object.

Your JSON object must have these exact keys:
- "summary": A concise, plain English summary of the email (max 100 words)
- "tone": Classify the tone as one of: Angry, Urgent, Neutral, Curious, Happy, Professional, Request
- "priority": Classify the priority as one of: High, Medium, Low
- "suggested_reply": A professional, concise, and appropriate reply based on the email content and inferred intent

Example of expected JSON output:
{
  "summary": "Meeting cancellation due to unforeseen circumstances, proposing reschedule.",
  "tone": "Neutral", 
  "priority": "Medium",
  "suggested_reply": "Hi [Name], thanks for the update. Let me know when you're available to reschedule."
}

Ensure all output is valid JSON. Do not include any other text or formatting outside the JSON object.
        

Gmail Trigger → Clean Email → OpenAI Analysis → Parse Response → 
Branch (Success/Failure) → Save to Airtable → Telegram Notification
        

const output = [];

for (const item of $input.all()) {
  // Extract sender information
  const from = item.json.From;
  const subject = item.json.Subject;
  
  // Get email body - prioritize plain text
  let body = item.json.snippet || item.json.Body || item.json.body || "(no body found)";
  
  // Basic HTML cleaning if needed
  if (typeof body === 'string') {
    body = body.replace(/<[^>]*>?/gm, ''); // Remove HTML tags
    body = body.trim();
    // Remove common signatures
    body = body.split(/-- \n|Sent from my iPhone/i)[0];
  }

  output.push({
    json: {
      sender: from,
      topic: subject,
      content: body,
      gmailMessageId: item.json.id
    }
  });
}

return output;
        

try {
  // Extract AI response
  const aiResponseString = $input.item.json.output;
  
  // Parse JSON
  const parsedJson = JSON.parse(aiResponseString);

  // Return success with thread ID
  return {
    json: {
      isSuccess: true,
      summary: parsedJson.summary,
      tone: parsedJson.tone,
      priority: parsedJson.priority,
      suggested_reply: parsedJson.suggested_reply,
      threadId: $input.item.json.threadId
    }
  };

} catch (error) {
  // Handle parsing errors
  return {
    json: {
      isSuccess: false,
      error: error.message,
      originalAiOutput: $input.item.json.output,
      threadId: $input.item.json.threadId
    }
  };
}
        

const now = new Date(new Date().toLocaleString('en-US', { timeZone: 'Africa/Lagos' }));
const hour = now.getHours();
let greeting;

if (hour >= 5 && hour < 12) {
  greeting = "Good morning! ☀️";
} else if (hour >= 12 && hour < 17) {
  greeting = "Good afternoon! 🌤️";
} else {
  greeting = "Good evening! 🌙";
}

return [{
  json: {
    ...$input.item.json,
    greeting: greeting
  }
}];
        

<b>{{ $json.greeting }}</b>

Hope you're having a good day!

<b>{{ $json.safeSender }}</b> just sent you an email with the subject <i>{{ $('Code').item.json.topic }}</i>

{{ $('Parse & Validate AI Response').item.json.summary }}

Tone seems <b>{{ $('Parse & Validate AI Response').item.json.tone }}</b>, <b>{{ $('Parse & Validate AI Response').item.json.priority }}</b> priority.

<b>Here's my draft response:</b>
<i>{{ $('Parse & Validate AI Response').item.json.suggested_reply }}</i>

Should I send it, or would you like to adjust anything?
        

🚨 Kairon AI Agent Error!
An email could not be processed because the AI's response was not valid JSON.

Error Message: {{ $json.error }}

Original AI Output:
{{ $json.originalAiOutput }}
        

Telegram Trigger → Parse Command → Branch (Revision/Command) →
Route Actions → Execute Action → Update Database → Send Confirmation
        

const message = $input.item.json.message;
const chatId = message.chat.id;
const messageText = message.text;

// Check if this is a reply to a revision prompt
if (message.reply_to_message && message.reply_to_message.from.is_bot) {
    return {
        json: {
            isRevision: true,
            revisionInstructions: messageText,
            chatId: chatId,
            repliedToMessageId: message.reply_to_message.message_id.toString()
        }
    };
}

// Parse standard commands like "/approve recordId"
const commandText = messageText.split(' ').slice(1).join(' ');
const parts = commandText.split(' ');

if (parts.length === 2 && parts[0].startsWith('/')) {
    const action = parts[0].substring(1);
    const recordId = parts[1];
    
    return {
        json: {
            isRevision: false,
            isValidCommand: true,
            action: action,
            recordId: recordId,
            chatId: chatId
        }
    };
}

// Invalid command
return {
    json: {
        isValidCommand: false
    }
};
        

try {
  const aiResponseString = $input.item.json.output;
  const parsedJson = JSON.parse(aiResponseString);

  return {
    json: {
      isSuccess: true,
      summary: parsedJson.summary,
      tone: parsedJson.tone,
      priority: parsedJson.priority,
      suggested_reply: parsedJson.suggested_reply,
      threadId: $input.item.json.threadId
    }
  };

} catch (error) {
  return {
    json: {
      isSuccess: false,
      error: error.message,
      originalAiOutput: $input.item.json.output,
      threadId: $input.item.json.threadId
    }
  };
}
        

Perfect! I've updated the draft with your changes. ✨

Here's the revised version:

<i>{{ $('Parse & Validate Revised Reply').item.json.suggested_reply }}</i>

How does this look? Ready to send or need more adjustments?