n8n Changes the World: the session deck

Part 1: the basics, six words

Most courses skip these, which is why automation feels like magic.

LLM

A Large Language Model predicts the next word, scores every option, picks one and repeats. ChatGPT, Claude and Gemini are all LLMs built on the Transformer, from Google's 2017 paper "Attention Is All You Need".

API

How one app asks another for something. A waiter: you never walk into the kitchen, you order and the dish comes back. Behind one tap in a delivery app sit a maps API, a payment API, a messaging API and an email API.

API key

How a service knows it is you, and who pays. A secret sent with every request. Never share it, use one per project, set a spending limit.

Credentials

Where n8n keeps your keys and sign-ins. Save once, pick it from a list in every workflow. n8n encrypts them before saving.

JSON

The language apps answer in: labels and values. A form with three fields reaches n8n as three labelled values.

Trigger

What starts a workflow: a schedule, an app event such as a new email, a form, or a webhook from another system. Nobody presses start.

Tokens and the context window

• An LLM reads, writes and bills in tokens, pieces of words. Rule of thumb: 100 tokens is about 75 English words. Arabic usually takes more.
• The context window is how much it can hold at once: instructions, files, the whole chat so far, and the answer.
• When a chat gets too long the oldest part falls out. If a long chat starts going wrong, start fresh with a summary.

Two ways an LLM gets paid

People pay a flat monthly subscription to chat in the app: ChatGPT, Claude, Gemini. Apps and workflows pay per token through the API. n8n uses the API, so every AI step in a workflow is an API call made with a credential.

Part 2: work that runs itself

Most of us still carry AI's output by hand: prompt, copy, paste, send, track. AI did one step, you did the other four.

Every workflow is a trigger, steps and an action. In n8n each box on the canvas is an API call with a credential: you fill in a form and n8n writes the request for you.

Two templates from the open n8n AI Agents library show the pattern. Email triage: a new email arrives, AI sorts it and drafts a reply, the draft waits in Gmail and urgent ones ping Slack. Invoice PDF to Sheets: a PDF lands in Drive, AI reads the supplier, number, date and total, a duplicate check runs, and the row lands in the sheet.

Automation

Your rules decide every step. Same path every run. Example: every Sunday, email me my deadlines.

AI automation

Fixed steps with AI in one of them, for the messy part. Example: form in, profile card out.

AI agent

AI decides every step. It picks its tools and loops until the goal is done. Example: find jobs that fit me.

Part 3: build Future Me 2030 yourself

Four checkpoints on a free n8n Cloud trial. No credit card, and the AI step uses n8n's built-in Gateway credits, so there is no API key to copy.

1. 1

   Set up

   Sign up at app.n8n.cloud/register, verify your email, and create a new workflow. You should see an empty canvas with one + in the middle.

2. 2

   Checkpoint 1: the form

   Add an n8n Form Trigger (On form submission). Title it Future Me 2030 and add three elements: First name (text), Your field (dropdown), Your 2030 dream (textarea). Set Respond When to Workflow Finishes so the card shows on the next page.

3. 3

   Checkpoint 2: test it

   Click Execute step, fill in the form on the test URL with your own answers, and switch the output to JSON. Your words come back as labels and values.

4. 4

   Checkpoint 3: the AI step

   Add a Basic LLM Chain after the form with Source for Prompt set to Define below. Under it, add an Anthropic Chat Model, choose Use Gateway credits as the credential, pick Claude Haiku 5.5, and paste the prompt below.

5. 5

   Checkpoint 4: the answer page

   Add an n8n Form node after the chain. Set Page Type to Form Ending, On submission to Show Completion Screen, the title to Meet yourself in 2030, and drag the AI's reply into the Completion Message. Run it and your card shows on the page.

The prompt

Drag the labels from the left panel into the prompt and n8n writes the curly brackets for you.

You are a warm, honest career coach writing from 2030.

Name: {{ $json["First name"] }}
Field: {{ $json["Your field"] }}
Dream: {{ $json["Your 2030 dream"] }}

Write their profile from 2030:
1. A job title that doesn't exist yet.
2. A day in their life, in 3 short lines.
3. Three skills to start learning this month.
4. One small step for this week.

Under 150 words. Plain text, no markdown.
Realistic, kind, specific.

Want it in Egyptian Arabic? Add one line: "Write it in Egyptian Arabic."

Part 4: agents

A workflow follows a recipe. Ask it to find three real jobs that match your 2030 card and it cannot: it only writes. An agent gets a goal and a toolbelt, then picks the steps itself.

An agent is an AI model with tools, memory and a loop: think, act, observe, decide, and repeat until the goal is met. It searches, compares and saves, then stops for your approval before anything goes out. The deck closes on a live data analyst agent built in n8n: ask in plain words, and it reads the database, writes and runs a read-only query, then draws the chart and explains it.

Twelve words, defined

LLM

A model that predicts the next token.

Token

A piece of a word. What it reads, writes and bills.

Context window

How much it can hold at once.

Model vs app

GPT is the engine. ChatGPT is the car.

API

How one app asks another for something.

Endpoint

Where the request goes, with GET or POST.

JSON

Labels and values. What apps answer in.

Credential

Proof of who is asking, and who pays.

API key

A secret string. Treat it like cash.

OAuth

"Sign in with Google." Approve once, revoke any time.

Trigger

Schedule, app event, form or webhook.

Webhook

Another app rings yours, instantly.