You already know how to connect LLMs to your code and applications. But what if you add automation to the mix? Imagine: you enter data into a spreadsheet — and the model instantly writes a summary, creates a task in Notion, or sends a reminder in Telegram. Without a single line of code. That’s the power of automation.

In this lesson, you’ll learn how to connect GPT with familiar tools: Google Sheets, Notion, and visual workflows in Make and Zapier. Everything that used to require manual work — can now be done by your personal AI assistant.

📚 What You’ll Learn

• How to use LLM prompts inside Google Sheets and Notion
• How to trigger the model automatically based on an event
• How to build “if A, then prompt B” workflows
• How to automate reports, summaries, reminders, and planning

📓 Key Terms

• Make / Zapier – Visual automation builders: connect apps together
• Webhook – A way to automatically send data from one system to another
• Trigger → Action – Standard structure: an event triggers an action
• Prompt – A request to a language model (LLM), can be part of an action
• API Token – A key to access an external service (e.g., OpenAI or Notion)

📊 Auto-Summaries and Reports in Google Sheets

When you have a spreadsheet — whether tasks, ideas, meetings, or notes — you can turn it into a live source of automatic reports. Instead of manually copying and analyzing, you connect ChatGPT via Make and get auto-summaries, priorities, and even daily reports.

💡 Why This Matters

Before: you open the sheet, read each task, write summaries manually, plan deadlines, and set priorities. Now: the model does all of this automatically.

📎 Analogy: like having an assistant who monitors your sheet and every time a new row appears — instantly writes a summary, highlights important points, and suggests a plan.

🧪 Example Automation: Task Summary

Suppose you have a table:

Scenario in Make:

1. 🟢 Trigger: new row in the table.
2. 💬 Prompt: "Write a short summary of the task: {{Description}}"
3. 🤖 Request: sent to OpenAI.
4. ✍️ Result: model returns text → written into Task Summary.

🔁 Other Use Cases

🧠 Priority Setting

Prompt:

Rate the priority of the task "{{Description}}" from 1 (low) to 5 (critical).
Consider the date, importance of changes, and short-term value.

🕒 Deadline Suggestion

Prompt:

Suggest a reasonable deadline for the task: {{Description}}.
Consider that the release is in 10 days.

📬 Daily Summary

Scenario runs once a day and collects tasks added in the last 24 hours:

Write a short report on the following tasks: {{task list}}.
Highlight what is done, what is critical, and what needs attention.

The result can be sent to Slack, by email, or saved in Notion.

⚠️ Watch Out For

• Too long descriptions → trim to 300–500 characters.
• Lack of context → add clarifications in prompts: “in project X”, “marketing task”, “for the marketing team”.
• Too frequent triggers → if it runs too often, you may hit token or API limits.

🧠 Notion as an Assistant

Notion has long been more than just a notebook. Today it’s a knowledge base, task tracker, brainstorming space, and content planner. And if you connect it to LLM (e.g., via OpenAI API and Make), Notion becomes a true AI assistant that helps you think, plan, and make decisions.

📎 Analogy: If before you wrote an idea in a notebook and thought about it yourself, now you have an attentive editor beside you. You jot down the idea — and it instantly offers an article plan, relevant tags, and difficulty rating.

🤖 Scenario: AI Content Assistant

Suppose you have a Notion database “📝 Article Ideas”. Each note has fields:

Automation with Make:

1. 🟢 Trigger: new record or update in “Article Ideas”.

2. 📦 Collect data: title + short description.

3. 💬 Form prompt:

   Create an article outline for the title "{{title}}" and description "{{description}}".
   Return:
   1. Outline
   2. 5 keywords
   3. Approximate difficulty level (beginner / intermediate / advanced)
   
   

4. 🧠 Send to OpenAI: Make sends the prompt via API.

5. ✅ Update Notion record: Model’s answer fills in Outline, Keywords, Difficulty.

📎 Analogy: Notion as Editor + Manager

Imagine you drop an idea into an editorial Trello board. Previously, an editor would read it, call you, clarify, propose a plan, and schedule it. Now your digital assistant does it all — quickly, structured, no reminders or meetings.

🧠 Other Automations in Notion

• 🔍 Classification: Model auto-assigns notes to categories: “UX”, “Tech Debt”, “Growth Ideas”.

• 📈 Prioritization: Add a task, model assigns priority:

  Rate the priority of the task "{{description}}" from 1 to 5.
  
  

• 📬 Weekly Digest: Make runs every Friday:

  Create a digest of ideas and tasks from the past week. Highlight interesting and promising ones.
  
  

Result: a big “smart” post in Notion or email.

⚠️ Tips for Implementation

• Ensure good input data quality — the better described the idea, the better the output.
• Add an AI Status field — to mark if a note is already processed.
• Don’t hesitate to iterate: let AI draft, you edit, then AI refine.

⚙️ Make/Zapier: if A, then Prompt B

Make and Zapier are visual automation platforms working on “if event A, then do action B”. Such workflows are called no-code workflows, because you don’t need to code — you connect ready-made blocks. And most importantly: GPT can become part of this chain, doing intelligent processing.

📎 Analogy: Imagine a factory. Before: machines only copied, moved, or sent files. Now, in the middle, there’s an AI operator who reads, understands, and reformulates.

🔁 Example Automated Workflows

🗓 Google Calendar → Meeting Agenda

Scenario:

1. A new meeting is created in Calendar.

2. Make pulls title and description.

3. GPT gets prompt:

   Create a short meeting agenda from the description "{{event_description}}"
   
   

4. Result goes into the event’s “Description”.

📎 Outcome: Participants see what to discuss right away — no 15-minute warm-up wasted.

🧠 Notion → First Paragraph of Article

Scenario:

1. You add a new article idea in Notion.

2. Make detects it and fetches title + summary.

3. GPT gets prompt:

   Write the first paragraph of an article with title "{{title}}" and topic "{{summary}}"
   
   

4. Result saved into “Introduction” field.

📎 Outcome: No blank page syndrome — AI gets you started.

📩 Gmail → Report Analysis

Scenario:

1. Gmail receives an email with a PDF or table.

2. Make extracts the content.

3. GPT gets prompt:

   Read the report and summarize. Highlight 3 key findings and 1 recommendation.
   
   

4. Result saved to Google Docs or sent to Slack.

📎 Outcome: No need to read long reports — AI gives you the essence.

🧠 Prompts as Workflow Steps

Each “GPT step” in Make/Zapier usually looks like:

• You create a prompt template with variables ({{title}}, {{body}}, {{meeting_notes}})
• These are auto-filled from previous steps
• Model generates text → passed to the next system (doc, database, message)

📎 Analogy: Like inserting an analyst, editor, or assistant in the workflow — understands context and adapts info for the next step.

💡 Especially Useful For

• Content automation (drafts, summaries, descriptions)
• CRM & support (responses, client summaries)
• Education (cards, quizzes, explanations)
• Marketing (emails, headlines, value props)

❌ Common Mistakes

• Sending too much text → model can’t handle, trim input
• Unstructured prompts → write “instruction + data”
• No fallback logic → scenario should handle no-response cases
• Ignoring token/API limits → watch consumption
• Non-parameterized prompts → hard to maintain and change

❓ FAQ

Can I use GPT with zero code? Yes — via Make, Notion AI, or Google App Script templates.

How to decide when to trigger the model? Tie it to events: new row, new message, new task.

What about sensitive data? Follow data policies: don’t send personal data to open LLMs without encryption.

✅ Summary

Now you know how to turn ChatGPT into an automated assistant that works with Google Sheets, Notion, and Make:

• Writes reports, descriptions, summaries, and reminders
• Reacts to events and updates data in real time
• Saves you time and cuts routine

Automation + LLM is not magic — it’s a tool. The main thing is to design a scenario where it’s useful.