How to Draft email responses with Pipedream

pipedream.com > Workflows > New

Create new workflow

Go to pipedream.com and click New Workflow in the top right. You'll land on the workflow builder with an empty canvas. The trigger step is automatically created as the first node in your workflow.

1. 1Click the 'New Workflow' button
2. 2Select 'Start with a trigger' option
3. 3Name your workflow 'Email Response Drafter'

Trigger Step > Select App > Slack

Configure Slack trigger

Click on the trigger step and select Slack as your source app. Choose 'New Message Posted to Channel' as your trigger event. This fires instantly when someone posts to your designated channel without polling delays.

1. 1Click the trigger step
2. 2Search for 'Slack' in the app list
3. 3Select 'New Message Posted to Channel' trigger
4. 4Click 'Connect Account' to authenticate

Slack Trigger > Channel Configuration

Select monitoring channel

Choose which Slack channel to monitor for email pastes. Create a dedicated channel like #email-drafts to keep this separate from general chat. The workflow will trigger every time someone posts in this channel.

1. 1Select your target channel from the dropdown
2. 2If channel doesn't exist, create it in Slack first
3. 3Leave 'Include Bot Messages' unchecked
4. 4Click 'Test Trigger' to verify connection

Workflow > Add Step > Node.js

Add filter step

Add a custom Node.js step to filter out short messages and bot posts. This prevents the workflow from firing on casual chat messages that aren't email content. You want to process only substantial text that looks like forwarded emails.

1. 1Click the + button below your trigger
2. 2Select 'Run custom Node.js code'
3. 3Name the step 'Filter Email Content'
4. 4Paste the filter logic in the code editor

Workflow > Add Step > OpenAI

Connect OpenAI account

Add an OpenAI step and connect your API account. You'll need an OpenAI API key from platform.openai.com. The GPT-4 model works best for professional email responses but costs more than GPT-3.5.

1. 1Click + to add another step
2. 2Search for 'OpenAI' in the app directory
3. 3Select 'Chat' action
4. 4Enter your OpenAI API key when prompted

OpenAI Step > Message Configuration

Configure GPT prompt

Set up the system prompt to generate professional email responses. Reference the Slack message text using steps.trigger.event.text. The prompt should specify your company tone, response length, and any specific instructions for your industry.

1. 1Set Model to 'gpt-4' for best results
2. 2Set Max Tokens to 500 for email-length responses
3. 3Configure the system message with response guidelines
4. 4Map the user message to the Slack text content

Workflow > Add Step > Node.js

Add response formatting

Create another Node.js step to format the AI response for Slack. This cleans up the GPT output, adds context about the original message, and formats it nicely for your team to copy-paste.

1. 1Add another custom Node.js code step
2. 2Name it 'Format Response'
3. 3Reference the OpenAI response using steps.openai.choices[0].message.content
4. 4Add formatting for Slack display

Workflow > Add Step > Slack > Send Message

Send response to Slack

Add a Slack Send Message step to post the drafted response back to the channel. Configure it to reply in a thread so responses stay organized. Include the original poster's username so they know their request was processed.

1. 1Add a Slack 'Send Message to Channel' step
2. 2Select the same channel as your trigger
3. 3Set Thread TS to steps.trigger.event.ts for threading
4. 4Map your formatted response to the message text

Workflow > Deploy > Test

Test the workflow

Deploy your workflow and test it by pasting a sample client email into your monitoring channel. The response should appear within 5-10 seconds as a threaded reply. Check that the tone matches your brand and the content addresses the original email appropriately.

1. 1Click 'Deploy' in the top right
2. 2Go to your Slack channel
3. 3Paste a sample client email
4. 4Wait for the threaded response

Node.js Steps > Error Handling

Configure error handling

Add a try-catch block to handle OpenAI API errors gracefully. When the API is down or rate limited, post an error message to Slack instead of failing silently. This keeps your team informed when the automation isn't working.

1. 1Edit your formatting step
2. 2Wrap the OpenAI call in try-catch
3. 3Add fallback message for API errors
4. 4Test with invalid API key to verify error handling

This code filters incoming Slack messages to process only email-like content and formats the AI response with proper threading. Paste this in your filter step to avoid triggering on casual chat messages.

export default defineComponent({
  async run({ steps, $ }) {
    const message = steps.trigger.event.text;

export default defineComponent({
  async run({ steps, $ }) {
    const message = steps.trigger.event.text;
    const user = steps.trigger.event.user;
    
    // Filter out bot messages and short content
    if (steps.trigger.event.bot_id || message.length < 100) {
      $.end('Message too short or from bot');
    }
    
    // Check for email-like patterns
    const emailPatterns = [
      /@\w+\.\w+/, // email addresses
      /from:/i,     // email headers
      /subject:/i,  // subject lines
      /dear|hi|hello/i // greetings
    ];
    
    const hasEmailPattern = emailPatterns.some(pattern => 
      pattern.test(message)
    );
    
    if (!hasEmailPattern) {
      $.end('Does not appear to be email content');
    }
    
    // Format for OpenAI processing
    const promptContext = {
      original_message: message,
      requested_by: user,
      timestamp: steps.trigger.event.ts,
      instructions: "Draft a professional, empathetic response that addresses the client's concerns directly. Keep it under 150 words and match our friendly but professional tone."
    };
    
    return promptContext;
  }
});