# How n8n works: a plain-English guide for marketers

> What n8n is, how workflows, nodes, items and executions fit together, and which marketing jobs suit it. Verified against n8n docs, Nov 2026.

Source: https://mentionx.ai/learn/n8n/n8n-how-it-works-marketers  
Author: Dipuck Jones, founder of MentionX  
Updated: 4 November 2026  
Level: Practitioner

**n8n is a workflow automation tool: you connect triggers, apps, AI models and logic on a visual canvas, and it runs those steps for you on a schedule or when something happens. For marketers it's a way to stop copying data between tools and to put AI to work inside repeatable processes.**

> **Key takeaways**
>
> - A workflow is a chain of nodes that starts with a trigger and passes items from step to step.
> - n8n Cloud prices by executions, and an execution is one full run however many steps it has.
> - The HTTP Request node lets you reach almost any API, even without a dedicated node.
> - Draft edits auto-save; nothing runs in production until you publish.

*Checked 4 November 2026 (CET) against n8n documentation and the n8n pricing page. n8n ships updates often and node options, plan limits and prices change, so confirm in your own instance and on the pricing page. "Our view" marks MentionX opinion. Ledgerline is a fictional company used for illustration.*

## What 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.

## How 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.

## What kinds of nodes are there?

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

| Family | Examples | Marketing use |
|---|---|---|
| Triggers | Schedule Trigger, Webhook, Form, RSS Feed Trigger, app triggers | Start a workflow at 08:00 on Mondays, or when a demo form arrives |
| App nodes | Google Sheets, Slack, HubSpot, Google Analytics, Google Ads | Read and write the tools you already use |
| Core nodes | HTTP Request, IF, Switch, Merge, Code, Remove Duplicates, Loop Over Items | Shape and route data, call any API |
| AI nodes | AI Agent, Basic LLM Chain, Text Classifier, Information Extractor, model sub-nodes | Summarise, classify, extract and decide |

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.

## Where 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.

## Drafts, 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.

> **Pro move:** add a sticky note to each workflow saying who owns it, what it does and what to do if it fails. Six months later, that note is worth more than the workflow name.

## How 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.

## Which 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.

| Job | Good fit? | Why |
|---|---|---|
| Copy new form leads into the CRM and alert sales in Slack | Yes | Clear trigger, clear destination, easy to test |
| Enrich leads with company data and route by segment | Yes | HTTP Request plus an IF or Switch node handles it |
| Draft a weekly performance summary from GA4 and Google Ads | Yes | Schedule, fetch, summarise with an AI node, post |
| Generate blog drafts from approved briefs | Yes, with review | Use a human approval step before anything is published |
| Run the same buyer prompts through AI assistants each week and log brand mentions | Yes | Schedule, HTTP or AI nodes, a data table and a Slack digest |
| Decide your positioning | No | A strategic judgement, not a workflow |
| Post to social accounts without anyone reading the copy | No | Brand risk outweighs the time saved |
| Replace your analytics tool | No | n8n moves and transforms data; it isn't a reporting database |

## The 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.

## How 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.

## Worked 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.

## Try 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.

## Quick 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.

## What 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](https://mentionx.ai/learn/n8n/n8n-cloud-vs-self-hosted). Or return to the [n8n for Marketers track](https://mentionx.ai/learn/n8n/n8n-for-marketers).

## Common 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.

## Sources

- n8n, [Plans and pricing](https://n8n.io/pricing/)
- n8n Docs, [Work with nodes](https://docs.n8n.io/build/understand-workflows/workflow-components/work-with-nodes)
- n8n Docs, [Save and publish workflows](https://docs.n8n.io/build/understand-workflows/save-and-publish-workflows)
- n8n Docs, [Track usage with Insights](https://docs.n8n.io/administer/observe-and-log/track-usage-with-insights)
- n8n Docs, [HTTP Request node](https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-base.httprequest)
- All accessed 4 Nov 2026.

## Frequently asked 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.

---

Written and reviewed by Dipuck Jones for MentionX Learn. The HTML version of this page: https://mentionx.ai/learn/n8n/n8n-how-it-works-marketers