From the third automation running productively, Zapier's bill regularly tips into the three-digit monthly range, and from the first serious productive setup with several workflows, it is often in the four-digit range. With the self-hosted version of n8n, it remains at around 25 euros plus infrastructure costs. This gap makes n8n the obvious choice for SMEs in 2026, provided that someone sets up the first workflows properly.
This article describes seven specific automations that typically deliver the fastest return when an SME plans its first n8n workflow. Each can be implemented in one to three days; the introductory quantity structure shows why the first workflows are usually amortised within four weeks. Before that, we briefly clarify what n8n is, what it costs and where it doesn't belong. Without this delimitation, tool decisions in SMEs regularly go wrong.
What n8n is - and what it isn't
n8n (pronounced "n-eight-n", short for "nodemation") is a workflow automation platform from Berlin, founded in 2019 by Jan Oberhauser. Functionally comparable to Zapier or Make: triggers receive an event (incoming email, webhook, schedule), a chain of nodes processes the data, which ends up in one or more target systems. You build these chains visually in the browser using drag-and-drop. Technically, each node can be extended with JavaScript or an experimental Python mode.
Three features distinguish n8n from the American competition. Firstly, the licence model: The Community Edition is available under the Sustainable Use License (Fair Code). You can operate it on your own infrastructure for an unlimited period of time and without a per-workflow fee, even in a commercial context, as long as you do not resell n8n as a hosted SaaS product yourself. Secondly, the choice of location: You host the instance where you want. With a German or European provider, the data protection discussion often takes ten minutes. Thirdly, the maturity level of AI workflows: With the 2.0 version from January 2026, LangChain is natively integrated. n8n itself mentions over 70 LangChain nodes for models, vector databases, embeddings and agent memory; in total, the project's GitHub repo lists "400+ native integrations" as the official status.
What n8n is not: a replacement for a workflow engine on the scale of a large corporation (BPMN modelling, audit trails according to banking standards, high availability across multiple data centres), not a replacement for a data integration platform such as Talend or Fivetran (the node editor is the wrong tool for million-line ETL), and not a replacement for a CRM, ERP or helpdesk. n8n connects these systems. It does not replace any of them.
Self-hosted or n8n Cloud: the two realistic ways
There are two sensible operating models for SMEs: the self-hosted Community Edition on your own server, or n8n Cloud (according to n8n support, hosted in the EU data centre, not officially guaranteed as a region guarantee).
The self-hosted version is free of charge with the licence, you only pay for the infrastructure. Realistically, it costs 8-30 euros per month for a small VM at Hetzner, IONOS or a regional hoster (after the Hetzner price adjustment in April 2026, the CPX22 will start at just under 8 euros), plus the setup time of a developer, typically half a day to a day including Docker-Compose, Postgres, reverse proxy with TLS and backup. It makes sense if you process sensitive data (personnel files, accounting, customer data), the number of automations is set to grow or you already operate your own server infrastructure.
n8n Cloud starts at 24 euros per month (Starter, 2,500 executions), 60 euros for the Pro plan with 10,000 executions, and 800 USD (around 740 euros) for the Business plan with 40,000 executions, multiple environments and Git integration. SSO/SAML is only included in the Enterprise plan. Cloud makes sense if you want to try out n8n first, don't have your own IT team and can live with the data protection conditions of the cloud offering. The execution limits are less limiting in practice than they sound: A typical lead routing workflow counts one execution per lead, not per internal node.
Rule of thumb: Under ten workflows with moderate frequency, cloud is worth it. From ten workflows, sensitive data or several teams involved, self-hosting is almost always cheaper and cleaner.
7 n8n-workflows-that-pay-off-for-medium-sized-businesses
The following seven automations are arranged in the order in which they typically deliver the fastest return. Workflow 1 and 2 usually amortise within three to four weeks, the rest in four to six.
1. Incoming invoices from the mailbox to accounting
The most common initial use case. A mailbox rechnung@firma.de receives PDFs. An n8n workflow extracts the attachments, sends them through an OCR API (e.g. mindee, AWS Textract or a locally hosted model), extracts the invoice number, date, amount, VAT ID and service description, compares the supplier against the vendor list in DATEV, Lexware or sevDesk and creates a journal voucher record. Disputed cases (unknown supplier, amount discrepancy against purchase order) are sent to a Slack or Teams channel for clarification.
Typical time required per invoice in manual mode: 4-7 minutes. With 400 incoming invoices per month, that's 27-47 hours, or €1,200-2,100 per month at a standard industry internal accounting hourly rate of €45 (model assumption). In practice, the pre-entry rate for automation is typically 70-85%; the rest is processed manually. Setup with two imported supplier templates: around two days.
2. Lead routing from the web form to the CRM and to the correct slack channel
Inquiry forms typically run into a collection inbox from which a salesperson distributes daily. n8n takes over this step: The webhook receives the form, enriches the email domain with company data (Hunter, hunter.io, or Apollo, apollo.io, via API), decides on the responsible contact person based on the industry, number of employees and requested service, creates the lead in the CRM (HubSpot, Pipedrive, CentralStationCRM and most relevant systems have native nodes), and sends a Slack message to the right channel with an "Accept" button.
The leverage lies less in the hours saved than in the response time. The much-cited MIT study by James Oldroyd (2007, later summarised in the Harvard Business Review as "The Short Life of Online Sales Leads") shows: The likelihood of qualifying a B2B lead is 21 times higher if the initial contact is made within five minutes - compared to 30 minutes. Setup: half a day, provided the CRM has a decent API.
3. Alerting at stock levels, servers or threshold values
A recurring pattern in SMEs: some important value is maintained in Excel or a database, someone "looks at it regularly" until they forget. n8n is well suited for these gap fillers. Time-controlled query (every hour or so) to a data source, threshold check, notification via the channel that the recipients use anyway: Slack, Microsoft Teams, SMS via Twilio or a call via PagerDuty/Opsgenie.
Concrete examples from typical SME setups: minimum stock levels in the merchandise management system (Sage, weclapp, Xentral), disc utilisation on a file server, error rate in a Shopware order pipeline, response times of an API. The value of these workflows rarely lies in the hours saved, but rather in the follow-up costs avoided. An unnoticed out-of-stock on an A-item costs a multiple of the entire monthly licence; an unnoticed full disc run on the mail server loses two days of business.
4. Order confirmations from the shop to accounting and to the shipping service provider
Anyone who runs Shopware, Shopify or another shop knows the pattern: an order comes in, it has to be entered into DATEV (or the accounting department first via an export), it has to be entered into the shipping software (Sendcloud, Shipcloud, DHL business customer portal) and, ideally, the customer receives tracking information as soon as it has been dispatched. n8n bundles these steps behind a webhook from the shop.
The specific trigger is usually the checkout.order.placed event from Shopware 6 (or the corresponding webhook from other systems). The node chain: Retrieve order detail, send booking record to the accounting API, generate shipping label via Sendcloud, send tracking email via your own mail system, reduce stock quantity. What sounds like a simple workflow costs 3-8 minutes per order without automation; with 50 orders per day and a conservative time assumption, around 15 hours per week, significantly more with the upper value.
5. Monitoring contract terms and notice periods
A SaaS licence for 24 months at €3,500 per year that is extended by a further twelve months because the notice period was missed by two weeks is a typical mid-four-figure individual loss for SMEs. This is exactly where the workflow comes in: A table (Notion, Airtable, Google Sheets or a separate database) holds all contracts with suppliers, SaaS tools and service providers, including the term and cancellation period. n8n runs daily, checks which contracts can be cancelled in the next X days and sends an overview to the person responsible. For critical contracts, two optional escalation emails plus a calendar entry.
Setup: half a day including data model. The ROI of this workflow results from a single avoided auto-renewal per year. For a medium-sized SaaS landscape of 30-60 active licences, this is the statistically normal hit rate if the workflow did not exist beforehand.
6. Employee onboarding (accounts, rights, devices)
As soon as the HR department creates a new employee in Personio (or a comparable HR system), n8n starts a sequence: create a Google Workspace account, add them to the right groups, invite a Slack account, create an onboarding project in task management (Asana, ClickUp, Jira), open a task in the IT ticket system for issuing devices, and send a welcome email with registration information on day 1 in the morning. Structured, traceable, without the office manager working through checklists until 7 pm the day before.
The time required for the manual variant varies with the depth of the tool, typically 2-4 hours per new hire. For an 80-person SME with 12 new hires per year, this equates to 24-48 hours, and 60-120 hours for 30 new hires. In addition, there is the unmeasurable but real damage of a bad day 1 experience. Setup: one to two days, because this is where most internal conventions come into play (group mapping, roles, device pool).
7th daily briefing from three systems for management
A time-triggered workflow, Monday to Friday at 8 a.m., pulls key figures from three or four systems: Sales from the ERP, new leads from the CRM, open tickets from the helpdesk, free liquidity from the banking API. A compact overview is then sent to the management by email or Slack. A dashboard in Power BI or Metabase fulfils the same purpose, but experience has shown that it is opened less frequently than a push message.
The value of this workflow lies not in the hours saved by a single person, but in better decisions earlier in the day. This is the most difficult ROI to quantify. If you want to evaluate it honestly, observe for a quarter how often sentences like "I think sales were up last week..." are uttered in the weekly meeting. If these are eliminated, the workflow has fulfilled its purpose.
What "4 weeks amortisation" means mathematically
A conservative model calculation using the example of workflow 1 (invoice receipt). Setup costs two days, internally or externally estimated at 1,200 euros. Monthly infrastructure costs of 25 euros for the VM, plus 40 euros for mindee Invoice OCR for 400 invoices (0.10 euros per page, mindee standard rate). Saved working time: conservatively 30 hours per month at an hourly rate of 45 euros, i.e. 1,350 euros. Net savings in the first month: 1,285 euros minus pro rata setup costs. The break-even point is therefore reached within the first month. The full savings are realised from month 2 onwards.
In the case of a purely internal setup without external service provider hours, the break-even point is reduced to two to three weeks. With external implementation and handover to the internal team, it is typically four to six weeks, depending on how cleanly the source system (accounting, CRM) exposes its API. The calculation is not so clear for each of the seven workflows individually. From workflow three to four, however, at least the infrastructure is amortised without discussion, because several automations run on the same n8n instance.
Where n8n doesn't fit
A few honest limitations so that the decision doesn't backfire. n8n is visual, but not a no-code tool in the strict sense. If you don't have a developer's eye, you will reach your limits with JSON structures, HTTP headers, OAuth token refresh or error handling. For teams with no IT experience at all, Zapier is more didactically accessible, even if it will be more expensive in the medium term.
For very data-intensive pipelines (millions of lines per run, complex transformations), n8n is overstretched. This is where an ETL platform or a separate data pipeline code is needed. And there is one point where the licence comparison really changes: for processes that are subject to a high degree of regulatory verification (medical technology, pharmaceuticals, banks, notary's office), the on-board resources of the Community Edition are not sufficient for complete audit trails. This is a point that is often ignored in the initial discussion, but which really tips the scales in the subsequent implementation decision. The Enterprise licence solves this technically via versioned workflow history and external log flow, but shifts the cost comparison noticeably upwards compared to Zapier/Make.
And finally: n8n is a tool, not a consulting concept. If you don't understand the underlying process, you will automate a mess more quickly. The honest sequence is always the same: understand the process, simplify the process, then automate it. If you turn this around, you automate the worse version of a bad idea - only then much faster.
Conclusion: one tool, two clear decisions, seven useful applications
The n8n question in SMEs boils down to two answers. Firstly: Can the data leave the company? If no - and in accounting, HR and order pipelines this is the default - self-hosting is not "also an option", but the default. If yes, cloud is convenient and cheaper in the initial learning phase. Secondly, is someone with developer insight available in the team to set up the first two or three workflows? If so, the project typically goes live within a few weeks. If not, it is worth buying in this expertise externally. Experience shows that the handover to the internal team is successful from the third jointly built workflow onwards.
The seven workflows above are a sensible order to start with, not a complete list. As soon as the first two are running productively, the next ones will almost fall into place automatically because the infrastructure is in place. A good first step before deciding on a tool is to select a single process that is currently time-consuming and do a thorough calculation. If you want an external perspective on this, nextlevels has an overview of app and workflow development and a dedicated page on n8n technology. The article on five SME projects with ROI in 12 months is the right neighbouring reading material for strategic classification in the first digitisation projects.