Module 10 Lesson 4: Summarization & Extraction

Large Language Models are the world's best "Editors." They can take 5,000 words of "Chaos" and turn them into 5 bullet points or—even better—a clean JSON object that n8n can understand.

1. Automated Summarization

• The Use-case: You have a 1-hour Zoom meeting transcript.
• The Workflow: Get transcript -> Send to AI -> "Summarize this into 3 'Action Items' and a 2-sentence summary." -> Send to Slack.
• The Value: Everyone on the team knows what happened without watching the video.

---

2. Structured Data Extraction (JSON)

This is the "Holy Grail" of automation. If a customer sends a messy email: "Hey I want to buy 5 of the Blue widgets, my address is 123 Main St and my zip is 90210." A regular tool can't read that. But AI can.

The Prompt: "Extract the quantity, color, address, and zip from this text. Return your answer ONLY as a JSON object."

The Result:

{ "qty": 5, "color": "blue", "zip": "90210" }

Now, you can pass this JSON directly to HubSpot or Google Sheets!

---

3. Sentiment Analysis

Does this customer sound happy or angry?

• You can ask the AI to categorize the email: Positive, Neutral, or Critical.
• If Critical, you trigger an Urgent Alert (Module 8).

---

4. Format Enforcement (Output Parser)

In n8n, you can add an "Output Parser" node to your AI chain. This ensures that the AI actually returns valid JSON and doesn't add conversational filler like "Here is your JSON...".

---

Exercise: The Digital Editor

1. Create a "Test Email" text with several pieces of hidden info (Price, Date, Product).
2. Setup an n8n AI workflow to extract those 3 fields into a JSON object.
3. Why is it important to tell the AI: "Return ONLY the JSON. Do not explain your self."?
4. How would you use AI to "Summarize" your daily Google Calendar (Module 7) into a morning "To-Do" list?
5. Research: What is the "JSON Object" mode in the OpenAI node?

---

Summary

Extraction is the bridge from the "Messy World" of humans to the "Structured World" of computers. By using AI to clean and summarize your inputs, you turn your n8n workflows into high-speed interpreters that can handle any data source.

Next Lesson: The autonomous bot: Autonomous Agents: The n8n Agent Node.