How to Build Your First Prompt Chain in Make.com — Step by Step

The exact configuration — modules, settings, data mapping — to connect your first 3-step prompt chain and have it running in under 90 minutes.

Understanding prompt chaining is one thing. Actually building it in Make.com requires knowing the exact module configuration, how to pass data between steps, and how to handle the inevitable first failure. This guide walks through every step.

What You Are Building

A 3-step lead research and email chain: Prompt 1 researches a company and returns structured JSON. Prompt 2 analyses the JSON to identify pain points. Prompt 3 writes a personalised email using both previous outputs. The trigger is a new row in Google Sheets (column A = company name). The result lands in column B as a ready-to-review email draft.

Step 1: Set Up Your Trigger

Create a new Make.com scenario. Add a Google Sheets trigger: "Watch for new rows." Select your spreadsheet and sheet. Set the trigger to watch column A. Test with one row containing a company name — Make.com will pull it in as test data.

Step 2: Add the Research Module (Prompt 1)

Add an OpenAI "Create a Completion" module. Set model to gpt-4o. Temperature: 0 (important — you need consistent, structured output). In the prompt field paste the research prompt, mapping the Google Sheets value to the company name placeholder. Set max tokens to 400.

Step 3: Add the Analysis Module (Prompt 2)

Add a second OpenAI module. In the User Message field, type your analysis prompt and map the output of the first OpenAI module to the data placeholder. This is the key step — Make.com calls this "data mapping" and it is done by clicking the variable picker icon and selecting the previous module's output.

Set temperature to 0.2 here — you want analytical consistency but slightly more flexibility than pure classification.

Step 4: Add the Writing Module (Prompt 3)

Add a third OpenAI module. Map both the Prompt 1 output (company data) and the Prompt 2 output (pain points) into the user message. Set temperature to 0.6 — you want some variation and naturalness in the writing. Set max tokens to 200 (enough for an 80-word email with JSON wrapper).

Step 5: Write Back to Google Sheets

Add a Google Sheets "Update a Row" action. Map the email output from Prompt 3 to column B of the same row that triggered the workflow. Add an error handler: if any module fails, send yourself an email with the company name and the error message.

Testing and Iteration

Test with 5 companies you know well. For each output, ask: is the research accurate? Are the pain points specific to this company or generic? Does the email reference a real detail? If pain points feel generic, add more specificity to Prompt 2. If the email is too long, reduce max tokens or add a word limit to Prompt 3.