What is n8n: a complete guide to business automation (with real SMB examples)

n8n is a workflow automation tool: it connects the applications a company already uses — email, ERP, CRM, spreadsheets, WhatsApp, databases, AI models — and makes things happen on their own when an event occurs. It's pronounced "n-eight-n", short for nodemation: every step of the flow is a node, and nodes are wired together on a visual canvas, with no code needed for most cases.

The difference from a traditional automation is that n8n isn't tied to a single piece of software: it sits in the middle. It takes an input from one side, transforms it, queries other systems and writes the result where it's needed. For an SMB that means getting tools that were never designed to talk to each other to do exactly that — which is precisely where most office time is lost.

How it works in practice

Every flow starts with a trigger: an incoming email, a submitted form, a new row in a database, a schedule, a call from another system. From there the nodes follow: one queries the ERP via API, one filters data, one calls a language model to interpret free text, one generates a PDF, one sends a Slack or WhatsApp notification. Between nodes you can add conditions, loops over lists, error handling and human approval steps. When ready-made nodes don't cover the logic you need, you drop in a small block of JavaScript or Python: that mix of visual and code is what makes n8n suitable for real business processes, which are rarely linear.

n8n, Zapier or Make: the recurring question. Zapier is the simplest and the most expensive as volumes grow, because you pay per operation. Make has a polished visual editor and good value per volume. n8n has two advantages that matter in B2B: it's open source and can be installed on a company server or a European cloud — so customer data doesn't necessarily pass through an external service — and it has no per-operation cost, so a flow running ten thousand times a month doesn't blow up the bill. In exchange it takes more technical skill to install, update and secure. There's also a managed cloud version from n8n that removes that burden for a subscription fee.

Privacy deserves an extra line, because in Italian companies it's often the deciding factor. With n8n on your own infrastructure, the data the flow processes stays where you decide; the only point where it can leave is the call to an external AI model, which can be limited to the strictly necessary fields or replaced by a locally run model. That's a substantial difference from closed SaaS platforms, and it makes GDPR assessments considerably simpler.

Five automations I see working most often in SMBs

First, sales inbox triage: incoming requests are classified, enriched with the customer's history and routed to the right person with a draft reply ready. Second, accounts payable: supplier invoices arrive by email, get read, the extracted data lands in the ERP and anomalies are flagged. Third, quote preparation, which in repeat-supply industries takes response time from days to minutes. Fourth, syncing systems that don't talk — e-commerce, warehouse, accounting — with no more manual file exports. Fifth, reporting: every Monday morning last week's numbers arrive already aggregated, with nobody opening a spreadsheet.

Where n8n is not the right answer

If the process already lives inside a single piece of software that handles it well, adding an external automation layer only adds complexity. If the source data is messy — duplicate records, three versions of the price list, hand-filled inconsistent fields — automation amplifies the mess instead of fixing it. And if the flow requires a decision the company has never written down, the first job isn't technical: it's deciding the rule.

What it costs

The software itself, self-hosted, is free; the server it runs on, at SMB volumes, usually costs between 10 and 40 euros a month. The real cost is design: an automation across two or three systems — analysed, built, tested on real cases and shipped — typically lands between 2,000 and 6,000 euros; a flow with an AI agent that interprets free text and writes into the ERP rises to 4,000-12,000 euros. At SMB usage, recurring AI model spend stays in the range of a few tens of euros a month. The detailed breakdown is in the guide on what an AI agent costs.

The fastest way to know whether n8n is right for your company isn't to evaluate the tool: it's to list the three repetitive tasks that steal the most hours every week and ask, for each, whether the data they need already exists in a digital system. Where the answer is yes, automation is almost always worth it and pays back within months. Where the answer is no, the first step is a different one — and it's worth knowing that before buying anything.