Day 4: OpenAI API inside Zapier & Make – Smart Automation

Day1

prompt engineering

Day2

Zapier trigger/action

Day3

Make multi‑step

Day4

OpenAI API inside tools

Now you control the AI directly – cheaper, customisable, and you can use any prompt from Day1 inside your Zaps & scenarios.

🤖 OpenAI API – what it is (in plain English)

📌 API = robot waiter

An API is like a waiter in a restaurant. You (your automation) write an order (prompt + settings), the waiter takes it to the kitchen (OpenAI servers), and brings back the dish (AI response). OpenAI API lets you call GPT‑4 or GPT‑3.5 directly from Zapier, Make, or any code.

Why API instead of pre‑built modules? Pre‑built modules (like Day3) are limited. With your own API key you can use any model, control parameters (temperature, max tokens), and it's often cheaper for high volume. Plus you reuse your Day1 prompts exactly.

Trigger

Gmail / form

Your API key

from OpenAI

API call

prompt ➔ response

💼 What you can build (simple examples)

Summarise support emails

Long email ➔ one‑line summary in Slack

Multi‑language

Translate incoming email to English before logging

Auto‑categorise

"billing", "tech", "sales" → tag in Sheets

All using your Day1 prompts, now automated.

🔑 Step 0: Get your OpenAI API key (no code)

Create API key

1. Go to platform.openai.com/api-keys (sign up / login).
2. Click + Create new secret key, name it "Zapier" or "Make".
3. Copy the key (starts with sk-...) – store safely.

Simplify: This is like a password for your automations to talk to OpenAI. Never share it publicly.

⚡ Two ways to use OpenAI API

Use OpenAI (GPT) integration (native).
Paste your API key once.
Choose model, write prompt with fields.
Best for: quick AI actions in linear Zaps.

Use HTTP module (call OpenAI API directly).
Full control: headers, JSON body, temperature.
Can chain AI calls, parse complex responses.
Best for: multi‑step, conditional AI.

🔨 Build #1: Zapier – Summarise every new Gmail

Trigger: Gmail – New Email

Same as Day2. Choose label or inbox. Get sample.

Action: OpenAI – Create Completion

Add step, search "OpenAI". Connect your API key.

Prompt (from Day1): "Summarise this email in one short sentence: {{Body}}"
Model: gpt-3.5-turbo (cheap & fast). Max tokens 50.

Action: Google Sheets – Add row

Map: From, Subject, and the AI summary (from step 2).

Done. Every new email gets a summary in your sheet.

🌐 Build #2: Make – Translate email to English (HTTP module)

Trigger: Gmail – Watch emails

Like Day3.

HTTP module (Make it simple)

Add HTTP – Make a request. Configure:

URL: https://api.openai.com/v1/chat/completions
Method: POST
Headers:
  Authorization: Bearer {{your API key}}
  Content-Type: application/json
Body:
{
  "model": "gpt-3.5-turbo",
  "messages": [
    {"role": "system", "content": "Translate the following email to English."},
    {"role": "user", "content": "{{subject}} - {{body plain}}"}
  ],
  "max_tokens": 200
}
                        

Day1 connection: The system prompt is pure prompt engineering.

Parse JSON & use translation

Add a JSON – Parse module to extract choices[0].message.content. Then use that text in Sheets / Slack.

🧘 Complex parts – made simple

It's just a way to organise data. Like a form: { "name": "John", "age": 30 }. The API returns JSON; you parse it to grab the text.

In Make, use the JSON – Parse module with sample data from the API. It auto‑generates the structure – no manual coding.

Keep it secret. Both Zapier and Make encrypt it. Never put it in a public field.

In OpenAI dashboard, you can limit the key to certain API calls (e.g., only Chat Completions).

💰 Keep costs low & prompts sharp

Use gpt-3.5-turbo for 90% of tasks (cheaper, fast).
Set max_tokens low – e.g., 50 for summaries, 200 for translations.
Reuse your Day1 library: copy and paste prompts into the API calls.
Test with one email, then turn on automation.

Practice (choose your tool)

Zapier: Sentiment

Build a Zap: new Typeform entry → OpenAI (analyse sentiment: positive/negative) → log in sheet with sentiment.

Make: Keyword extract

Scenario: Gmail trigger → HTTP OpenAI → extract keywords (prompt: "list 3 keywords") → Slack DM with keywords.

Combine with Day3 router

Take Day3 scenario, replace built‑in AI with HTTP OpenAI module (so you control model/prompt exactly).

📚 Friendly resources

Day 4: You can now call the OpenAI API inside Zapier & Make

✔ Understood API = waiter taking order
✔ Got your own API key (safe & simple)
✔ Built Zapier summariser & Make translator
✔ Reused Day1 prompts in real automation
✔ Demystified JSON & HTTP calls