Connecting ChatGPT via Zapier & Make

Automation with Zapier and Make (Formerly Integromat)

ChatGPT becomes dramatically more powerful when you stop treating it as a manual tool and start integrating it into your automated workflows. Zapier and Make let you connect ChatGPT to hundreds of other apps — so AI runs automatically when triggered by events, without you lifting a finger.

The Core Concept: AI in Your Workflow

Without automation: You open ChatGPT, paste content, write a prompt, get the output, and do something with it. Repeat for every piece of content.

With Zapier/Make: A trigger fires (new email arrives, form is submitted, a row is added to a spreadsheet), ChatGPT processes it automatically, and the result goes somewhere useful — a Slack message, a CRM update, a draft in your email.

You design the workflow once. After that, AI runs in the background without you.

Zapier vs. Make

Zapier:

• More accessible — simpler interface, less technical knowledge required
• Better for simple linear workflows (trigger → step 1 → step 2 → action)
• Wider app library
• More expensive at scale

Make (formerly Integromat):

• More powerful — visual flow builder with branches, loops, error handling
• Better for complex workflows with multiple paths and conditions
• More affordable for high-volume automation
• Steeper learning curve

For most professionals starting out: Start with Zapier. Move to Make when you need more complex logic.

Setting Up ChatGPT in Zapier

1. In Zapier, create a new Zap
2. Choose a trigger (new email, form submission, Google Sheets row, Slack message, etc.)
3. Add an action step: search for "ChatGPT" or "OpenAI"
4. Connect your OpenAI API key (Settings → API → Create new secret key)
5. Choose "Send a prompt" as the action
6. Write your prompt using dynamic fields from the trigger

The key is using Zapier's dynamic fields to inject trigger data into your ChatGPT prompt.

Practical Automation Examples

1. Auto-Summarize Incoming Emails

Trigger: New email received in Gmail (filtered by sender or label) ChatGPT prompt:

Summarize this email in 3 bullet points. Include:
- What they want or are asking
- Any deadlines mentioned
- Required action on my end

Email subject: {{Subject}}
Email body: {{Body Plain}}

Action: Create a task in Notion/Asana/Todoist with the summary as the description

2. Customer Support Auto-Triage

Trigger: New support ticket in Zendesk/Intercom/Help Scout ChatGPT prompt:

Analyze this support ticket and respond with JSON:
{
  "category": "[billing|technical|feature_request|account|other]",
  "priority": "[urgent|high|medium|low]",
  "sentiment": "[frustrated|neutral|happy]",
  "summary": "One sentence description of the issue",
  "suggested_response": "A helpful first response to this customer"
}

Ticket:
{{Ticket Subject}}
{{Ticket Body}}

Action: Update the ticket with category/priority tags, save suggested response as internal note

3. Content Repurposing Pipeline

Trigger: New blog post published (via RSS feed or webhook) ChatGPT prompt:

A new blog post has been published. Create three pieces of social content from it.

Post title: {{Title}}
Post excerpt: {{Excerpt}}
Full post URL: {{URL}}

Generate:
1. A LinkedIn post (150-200 words, professional tone, ends with a question to drive comments)
2. Three Twitter/X threads (first tweet + 4 follow-up tweets, conversational)
3. A 60-word email newsletter teaser (for our weekly digest)

Return as JSON:
{
  "linkedin": "...",
  "twitter_thread": ["tweet1", "tweet2", "tweet3", "tweet4", "tweet5"],
  "email_teaser": "..."
}

Actions: Create drafts in Buffer/Hootsuite for LinkedIn and Twitter, add to email draft

4. Meeting Notes → CRM Update

Trigger: New note added in your meeting notes app (or Google Docs, Notion) ChatGPT prompt:

These are notes from a customer meeting. Extract CRM-relevant information.

Notes:
{{Meeting Notes Content}}

Return JSON:
{
  "company_name": "",
  "contact_name": "",
  "pain_points": [""],
  "budget_mentioned": "",
  "timeline": "",
  "next_steps": [""],
  "deal_stage": "[discovery|evaluation|proposal|negotiation|closed]",
  "follow_up_date": "YYYY-MM-DD or null"
}

Action: Update the relevant deal in Salesforce/HubSpot/Pipedrive

5. Job Application Screener

Trigger: New job application submitted via form (Typeform/Google Forms) ChatGPT prompt:

Screen this job application for a [Role] position.

Required qualifications: [list must-haves]
Nice-to-have qualifications: [list preferred]

Application:
Name: {{Name}}
Experience: {{Years Experience}}
Cover letter: {{Cover Letter}}
LinkedIn: {{LinkedIn URL}}

Evaluate and return:
{
  "recommendation": "advance|reject|maybe",
  "reasoning": "2-3 sentences",
  "required_qualifications_met": ["list which ones"],
  "missing_requirements": ["list any gaps"],
  "standout_factors": "anything exceptional"
}

Action: Add to screening spreadsheet, send "thank you for applying" email with delay based on recommendation

Structuring Prompts for Automation

When prompts run automatically (no human reviewing before submission), they need to be more structured:

Always request structured output (JSON): Downstream Zapier/Make steps need to parse the ChatGPT output. JSON is easier to work with than prose.

Include validation instructions:

If the input doesn't contain enough information to complete this task, 
return: {"error": "insufficient_data", "reason": "what's missing"}

Set explicit constraints:

Return ONLY valid JSON. No markdown code blocks, no explanatory text. 
Start the response with { and end with }.

Handle edge cases:

If the email is spam or automated (newsletters, notifications), 
return {"category": "skip", "reason": "automated email"} and stop processing.

Connecting to Your Data

Zapier and Make can pass more than just triggered data to ChatGPT:

Look up CRM data before prompting: Use a Salesforce/HubSpot "find record" step before ChatGPT to add customer history to the prompt.

Pull from Google Sheets: Use Sheets as a simple database — look up products, prices, or templates to include in your prompt.

Chain multiple AI calls: Use ChatGPT output as input to a second ChatGPT call for a two-step refinement.

Error Handling

Automated workflows break. Build in resilience:

• Add a "catch" path in Make for when ChatGPT returns an error
• In Zapier, turn on error notifications so you know when a Zap fails
• Test your prompts manually before going live — paste sample trigger data and verify the output
• Log responses to a spreadsheet so you can audit what happened

Cost Management

ChatGPT API costs money per token. For high-volume automations:

• Use GPT-4o-mini instead of GPT-4o for tasks that don't need deep reasoning
• Set max_tokens limits in your API call to prevent runaway costs
• Add a filter step before the ChatGPT step to skip low-value triggers

Next lesson: Building your prompt library — saving, organizing, and reusing your best prompts.