How n8n works: a plain-English guide for marketers

#01What does n8n actually do?

n8n moves data between tools and decides what happens to it on the way. A trigger starts the workflow: a form submission, a new row in Google Sheets, a schedule, a webhook from your website, or a chat message. Each node after that does one job, such as fetching data, filtering it, asking an AI model to classify it, writing to the CRM or posting to Slack.

According to the n8n pricing page, every plan includes unlimited users, unlimited workflows and every integration. You pay for how often workflows run, not how many you build. That pricing model suits marketing teams who want many small automations without counting seats.

#02How does data flow through a workflow?

Data travels as a list of items. If a Google Sheets node reads 40 rows, the next node receives 40 items and usually runs once for each. This is the single most useful idea in n8n, because it explains most surprises:

• A Slack node after a 40-row read will post 40 messages unless you aggregate first.
• An AI node after 40 items will make 40 model calls, which affects cost.
• A filter that removes 35 items leaves 5 for the rest of the workflow.

You can see the items at every node. Click a node after running it and the Output panel shows them as a table, raw JSON or a schema. When something looks wrong, the answer is nearly always visible in the output of the node before.

#03What kinds of nodes are there?

The n8n documentation groups nodes into a few families. You'll use all of them in this track.

App nodes cover common operations. When one doesn't support what you need, the n8n docs recommend using the HTTP Request node with the same stored credential.

#04Where do AI models fit?

AI nodes sit inside the same workflow as everything else. That's the point: a model is useful when it receives the right data and its output goes somewhere useful. A typical marketing pattern is fetch, think, act: fetch new leads, ask a model to classify them against your ideal customer profile, then route them in the CRM. Lessons 8 to 12 build these.

Our view: the biggest gains for marketers come from putting a small, well-defined AI step inside a reliable workflow, not from giving an agent broad freedom on day one.

#05Drafts, publishing and executions

n8n auto-saves your edits as a draft. According to the n8n docs on saving and publishing, publishing makes a workflow live and locks it to a version: production webhook URLs start working, schedules run and app events trigger it. Your later edits don't affect production until you publish again.

Each run is an execution. You can browse past executions, see the data at each node and re-run failures. Manual test runs from the editor are also executions, but the n8n Insights docs note that headline production metrics exclude manual and sub-workflow executions.

#06How this track is organised

Lessons 1 to 4 cover foundations. Lessons 5 to 7 build workflows with Sheets, Slack and the CRM. Lessons 8 to 12 add AI, from structured outputs to an AI visibility monitor. Lessons 13 to 15 cover monitoring, reporting and MCP. Lessons 16 to 20 make workflows reliable, secure and affordable.

#07Which marketing jobs suit n8n, and which don't?

n8n is good at work that is repetitive, rule-based at the edges and fuzzy in the middle. The fuzzy middle is where AI nodes now help. Use this table to sort your own ideas.

#08The vocabulary you'll meet in every lesson

• Workflow. A canvas of connected nodes. Each workflow has one or more triggers.
• Node. A single step. Trigger nodes start workflows; action nodes do something; core nodes (IF, Switch, Merge, Code, HTTP Request) shape data and control flow.
• Item. n8n passes data between nodes as a list of items. Ten new leads are ten items, and most nodes run once per item.
• Expression. A small snippet such as {{ $json.email }} that pulls a value from the incoming item into a field.
• Credential. Stored login details for a service. You create them once and reuse them across workflows.
• Execution. One full run of a workflow. n8n Cloud pricing is based on executions, and an execution counts once however many steps it has.
• Publish. Workflows auto-save as drafts. Publishing makes a version live, which turns on production webhook URLs and schedules.
• Cluster node. An AI root node (such as the AI Agent) with sub-nodes attached for the model, memory, tools and output parser.

#09How n8n compares with no-code automation tools

Marketers often arrive from simpler trigger-and-action tools. The differences that matter in practice:

• Branching and loops are normal. IF, Switch, Merge and Loop Over Items nodes let one workflow handle several paths. In simpler tools that often means several separate automations.
• The HTTP Request node opens any API. If a service has an API but no n8n node, you can still call it. Many node pages in the n8n docs point to the HTTP Request node for operations the node doesn't support.
• You can see the data at every step. Each node shows its input and output, which makes debugging far easier than reading a log.
• Pricing counts whole runs, not steps. A 20-step workflow costs the same execution as a 2-step one on n8n Cloud. That changes how you design: one richer workflow is often cheaper than several small ones.
• You can self-host. The Community edition is free to run on your own server for internal business use, under a fair-code licence covered in the next lesson.

The trade-off is a steeper start. You will meet JSON, expressions and occasional JavaScript. This track keeps code to short, copyable snippets.

#10Worked example: Ledgerline's first automation shortlist

Ledgerline is a fictional accounts-payable software company. Details are illustrative.

Ledgerline's two-person marketing team listed what ate their week. Copying webinar sign-ups from the webinar tool into the CRM took about an hour a week and often went wrong. Posting a Monday performance summary in Slack took half a morning. Checking whether AI assistants mentioned Ledgerline when asked "best accounts payable software" was something nobody had time for.

They scored each task for frequency, error cost and data access. Webinar sign-ups scored highest: frequent, error-prone and fully available through APIs. The Monday summary came second. The AI mention check came third because it needed API keys and a decision about which prompts to track.

They built the sign-up sync first, as a Webhook trigger into a HubSpot node with a Slack alert. It was a five-node workflow. The point wasn't the hour saved, it was learning how items, credentials and publishing work on something low-risk before touching AI.

Walkthrough

1. Sign up for an n8n Cloud trial or open your company's instance. The trial on the pricing page allows 1,000 executions over 14 days, which is plenty for this track's exercises.
2. Open Overview and create a workflow. Name it "Sandbox - your name".
3. Click Add first step and choose Trigger manually. This is a manual trigger for testing.
4. Add an Edit Fields (Set) node. Add a field channel with the value newsletter and a field topic with the value AI search.
5. Click Execute workflow. Open the Set node and look at the Output panel in Table, JSON and Schema views.
6. Add a second Set node. In a field, type {{ $json.topic }} weekly and watch the preview resolve the expression.
7. Look at the Executions tab. Manual runs appear there too, depending on your workflow settings.

#11Try it yourself

Pick three repetitive marketing tasks from your last month. For each, write one line for the trigger ("when a form is submitted"), the steps, and the destination. Score each 1-3 for frequency, error cost and data access. Keep the highest-scoring one: you'll build it during this track.

#12Quick quiz

1. What is an item in n8n?

Answer: One unit of data passing between nodes. Most nodes run once for each item they receive.

2. Does a workflow with 15 steps use 15 executions on n8n Cloud?

Answer: No. The pricing page says an execution is one full run, regardless of the number of steps.

3. What does publishing do?

Answer: It makes a version of the workflow live, so production webhooks and schedules run on that version.

#13What should you check before the next lesson?

• You can create a workflow and run it manually.
• You can explain items to a colleague.
• You have a shortlist of three tasks to automate.

Next: n8n Cloud vs self-hosted: plans, editions and licence. Or return to the n8n for Marketers track.

#14Common questions

Is n8n open source? Not in the OSI sense. It uses the Sustainable Use License, which n8n calls fair-code. Lesson 2 explains what that allows.

Do I need to code? No for most workflows. Expressions and the occasional Code node help, and this track gives copyable snippets.

Can n8n call AI models? Yes. It has AI nodes for models, agents, classifiers and extractors, covered in lessons 8 and 9.