How to Add AI to Your Make (Integromat) Automation Workflows
Make.com is the most powerful no-code automation platform available — and when you add AI to your Make scenarios, the workflows go from automated to intelligent. Here is exactly how.
The Combination That Changes Everything
Make handles the data movement. AI handles the thinking. Together they replace entire categories of manual work.
Traditional Make automations move data between apps according to fixed rules: if this, then that. They are fast and reliable but they cannot interpret, classify, summarise, or generate. They break the moment input data does not match the expected pattern.
Adding AI to Make scenarios gives your automations the ability to read and understand unstructured text, make judgment calls about what to do with data, generate content based on context, and adapt to variation in input — transforming brittle rule-based workflows into genuinely intelligent systems.
Using Make’s Native OpenAI Module
Make has a built-in OpenAI module that connects to the Chat Completions API without any custom HTTP configuration.
Add the OpenAI module to your scenario
In your Make scenario, click the + button to add a module. Search for ‘OpenAI’ and select it. Choose ‘Create a Chat Completion’ as the action. Connect your OpenAI account by adding your API key in the connection setup.
Configure the model and prompt
Select your model (gpt-4o-mini for speed and cost, gpt-4o for quality). Add a System message with your role definition and instructions. Add a User message that includes the dynamic data from earlier modules using Make’s variable picker.
Map the response to downstream modules
The OpenAI module outputs the AI’s reply as a text variable. Map this output into the next module — whether that is updating a CRM record, sending an email, creating a document, or posting to Slack.
Handle token limits and errors
Set max_tokens appropriate to your use case. Add an error handler on the OpenAI module that catches 429 rate limit errors and retries after a 60-second delay. Log failed executions to a Google Sheet or Airtable for review.
Using the HTTP Module for Advanced Control
For more control over the API call — custom headers, response format, streaming — use Make’s HTTP module to call OpenAI directly.
Add an HTTP Make a Request module. Configure:
HTTP Module Configuration
- URL: https://api.openai.com/v1/chat/completions
- Method: POST
- Headers: Authorization: Bearer YOUR_KEY, Content-Type: application/json
- Body type: Raw / JSON
- Parse response: Yes (enables dot-notation access to response fields)
The response body is then accessible as data.choices[].message.content in subsequent modules. Using the HTTP module gives you access to parameters the native OpenAI module does not expose — including response_format: {type: "json_object"} for structured output.
📌 Always set Parse response to Yes. Without it, the entire response is a text string and you cannot access individual fields using Make variable mapping.
Chaining Multiple AI Calls in One Scenario
The most powerful Make + AI patterns involve multiple AI calls in sequence — each building on the previous result.
Classify then Act
First AI call: classify the input (support ticket category, lead quality score, document type). Second AI call: generate the appropriate response or action based on the classification. Each call is simpler and more accurate than trying to do both in one prompt.
Extract then Enrich
First AI call: extract structured data from unstructured input (name, email, company from a business card photo). Second AI call: enrich the extracted data (research the company, generate an ICP score, suggest a personalised opening line for outreach).
Generate then Validate
First AI call: generate content (email draft, product description, meeting summary). Second AI call: critique the generated content against specific criteria (tone, accuracy, length, required inclusions). Only proceed if the validation passes.
Copy-and-Adapt Workflow Patterns
| Scenario Name | Trigger | AI Role | Output |
|---|---|---|---|
| Support ticket triage | New email to support@yourcompany.com | Classify category + urgency + draft reply | Tagged ticket + draft response in helpdesk |
| Lead enrichment | New contact added to CRM | Score against ICP + generate company summary | Enriched CRM record + Slack alert for hot leads |
| Content repurposing | New blog post published | Generate LinkedIn post + 3 tweet variants + email snippet | Drafts created in Buffer queue for review |
| Invoice processing | New PDF attachment received by email | Extract vendor, amount, date, line items | New record created in accounting tool |
| Meeting follow-up | Otter.ai transcript ready webhook | Extract action items + decisions + generate follow-up email | Email draft sent to meeting organiser for review |
| Weekly KPI digest | Scheduled: every Monday 8am | Analyse metrics data + generate narrative insights | Formatted digest posted to Slack #leadership |
Making AI Scenarios Reliable
Error handling
- Always add error handlers on AI modules — API calls fail more than database operations
- Set up Make’s built-in email alerting for scenario failures
- Log every AI call input and output to a data store for debugging
- Use Make’s Ignore error handler for non-critical AI enrichment steps
- Add a fallback value when AI returns empty — never let downstream modules receive null
Cost control
- Set max_tokens on every API call — uncapped calls are a budget risk
- Use gpt-4o-mini for classification and extraction, gpt-4o only for quality-critical generation
- Monitor OpenAI usage dashboard weekly and set spend alerts
- Cache results for identical inputs using a Make data store lookup before calling the API
- Process items in batches during off-peak hours for non-urgent workflows
Want Make + AI Workflows Built for Your Business?
SA Solutions designs and builds production Make.com scenarios with AI at every intelligent decision point. We handle architecture, error handling, cost management, and monitoring.
