Process automation on your infrastructure – with n8n and custom ETL pipelines

Order processing, approvals, data flows between ERP and CRM, reporting: we automate the business processes your people handle by hand every day. With n8n on your EU servers and custom code where visual tools stop – no US cloud, and pricing out in the open. AI automation is added where it measurably takes work off your team's plate, not as a label.

Most automation vendors sell you either a Microsoft licensing estate or a SaaS subscription where your order data travels through someone else's data centers and every workflow run is billed individually. We take the other path: n8n – an open-source workflow engine with no license costs – runs on your infrastructure, set up, connected to your ERP, CRM and accounting, and maintained by us. Where a visual tool is not enough, say with legacy systems that have no API or with large data volumes, we build custom ETL pipelines in TypeScript. What that can add up to shows in real users' numbers: Delivery Hero saves around 200 hours per month through automated ops workflows alone, according to n8n, the vendor. And manually processing a single incoming invoice costs €15–40 according to the Billentis study – automated, process costs drop to a fraction. The workflows and every line of code belong to you; if we ever part ways, everything keeps running.

When automation is the wrong answer: if a process doesn't run cleanly even when done manually – unclear responsibilities, constant exceptions, nobody knows the target sequence – automating it only makes the chaos faster. Clean up first, then automate; we will tell you exactly that in the process check. Automation is equally pointless for one-offs without repetition: a sequence that happens three times a year doesn't justify building a workflow. And if you want to connect two standard SaaS tools with a few hundred runs a month and no sensitive data involved, Make or Zapier is often all you need – that's an afternoon of configuration, not a project. We will tell you honestly which case you are in before you spend any budget.

Logo von 20moments
Logo von Bemer
Logo von Berliner Volksbank
Logo von Clinomic
Logo von Consentmanager
Logo von Deloitte
Logo von DFDC
Logo von einhorn
Logo von Cushman & Wakefield
Logo von in-motion
Logo von myticket
Logo von TES H2

Three terms that come up in every automation conversation – and what they actually mean, so you can compare proposals.

Definition

Process automation
The technical execution of recurring business processes without manual in-between steps: an order automatically triggers invoicing, inventory posting and customer notification, instead of someone copying data from system to system. The payoff lies in processes with high repetition rates and clear rules – manually handling an incoming invoice costs €15–40 per document according to the Billentis study; automated, process costs drop by up to 70%.

Definition

Workflow engine (n8n)
Software for workflow automation: triggers (a new order, an incoming email, a schedule) start a chain of steps across interfaces to your systems. n8n is an open-source workflow engine with several hundred ready-made integrations plus the ability to embed your own code. The Community Edition is free of license costs and runs on your own infrastructure – we set it up there, connect your systems and look after the operation. Your data never leaves your servers.

Definition

ETL pipeline
Extract, transform, load: individually programmed data routes that pull data from one system, reshape it and write it into another – say, nightly product-data synchronization between ERP and shop, or consolidating sales figures for reporting. ETL pipelines written in code (in our case: TypeScript on EU servers) win where visual tools hit their limits: legacy systems without an API, direct database access, large data volumes, complex transformation logic.

This page covers business processes: order processing, approval chains, data flows between ERP and CRM, automated reporting. Marketing workflows – lead funnels, newsletter logic, CRM campaigns – are built with the same tools; the focus here is on the business processes behind them. If your workflows grow into the need for a full application with its own interface, roles and permissions, you are looking at custom software development. And if you want to know what is behind the buzzword of AI agents that supposedly run processes on their own: our guide to agentic AI separates what works from the marketing.

Where automation gets concrete for trade contractors and mid-sized companies

Abstract automation promises help no one. So let's be specific: these are the processes we encounter most often in larger skilled-trades businesses (electrical, plumbing, construction trades) and in mid-sized companies – the German Mittelstand – and where the first workflow pays for itself fastest.

Larger skilled-trades businesses

Quote and order intake: Inquiries from email and web forms land in your industry software in structured form – with automatic follow-up reminders so no job dies in an inbox. The workflow requests missing documents on its own.

Incoming invoices and accounting handover: Supplier invoices are read from the inbox, matched to the right project, routed for approval and prepared for accounting – all the way to an export for DATEV, the accounting standard in Germany. At €15–40 in process costs per manually handled document (Billentis), this is often the process with the shortest path to payback.

Site and project documentation: Photos, measurements and reports from the field are automatically filed with the right project, instead of being typed up at the office after hours.

Maintenance and recurring jobs: Service contracts generate their own appointments, customer reminders and report templates – dispatch only sees the exceptions.

Manufacturing and trading SMEs

ERP–CRM synchronization: Customer, product and order data is maintained once and kept in sync automatically – the classic starting point for most of our projects.

Supplier price lists: CSV and SFTP imports run overnight instead of someone keying in prices by hand – including a variance report waiting in the morning.

Sales and management reporting: Figures from ERP, shop and accounting are consolidated and ready every morning, without anyone stitching exports together.

The most common objection we hear from the trades: "Our industry software has no interface." That is exactly what custom ETL routes are for: direct database access, CSV exchange over SFTP, structured email processing if it comes to that. A missing API is a detour, not a roadblock – and we verify that in the process check before you spend any budget.

Security & compliance

Your customers trust you with sensitive data. We take that responsibility seriously.

Whether it's your data protection officer, your IT security team or a major customer's procurement: when security questions come up, you have solid answers. We build in data protection and security from day one and follow the principles of the Twelve-Factor App. That makes approvals and audits easier and gives you the technical foundation to meet requirements under GDPR, ISO 27001 or HIPAA.

Frameworks whose technical requirements we implement in our projects

  • GDPR
  • ISO 27001
  • HIPAA
  • NIS2
Discuss your requirements in an intro call

Three ways to automate processes – and when each one wins

Anyone offering you only one path is selling you their business model, not your solution. There are three serious ways to automate recurring workflows – and the main difference is not feature count but data control and how costs behave as you grow. All third-party prices: as of 18 August 2026.

Data control

SaaS iPaaS (Make, Zapier)
Your process data runs through the vendor's cloud, usually with US ties; GDPR compliance only achievable via data-processing agreements and standard contractual clauses
n8n on your infrastructure
All data stays on your EU servers (e.g. Hetzner); no US-cloud transfer, a data-processing agreement exists only with us
Custom ETL pipeline (code)
Full control: runs in your environment, on-premises if you want; suitable even for data that must never leave the building

Costs as volume grows

SaaS iPaaS (Make, Zapier)
You pay per task or operation: Zapier from $19.99/month for 750 tasks, Make from $9/month for 10,000 operations – and Make counts every module action individually, not the workflow run. Costs rise linearly with every run
n8n on your infrastructure
License costs: €0 (Community Edition, Sustainable Use License). Server costs €20–50/month, high availability €150–400. Whether 500 or 50,000 runs a month: operating costs stay nearly flat
Custom ETL pipeline (code)
No license costs, but upfront development effort. Pays off with high data volumes and long-lived core processes – after that, the pipeline runs for bare server costs

Integration depth (ERP/CRM)

SaaS iPaaS (Make, Zapier)
Ready-made connectors for common SaaS tools; with German mid-market ERP systems, in-house builds or on-premises software, the catalog runs out fast
n8n on your infrastructure
Several hundred integrations plus your own code steps: REST, SOAP, direct database access, legacy CSV/SFTP routes – even an ERP without a modern API is reachable
Custom ETL pipeline (code)
As deep as needed: direct database integration, proprietary interfaces, batch processing of large datasets, your own error and restart logic

Maintenance and operations

SaaS iPaaS (Make, Zapier)
The vendor maintains the platform; in return, connectors and pricing can change at any time, and during outages you wait in line with everyone else
n8n on your infrastructure
Updates and monitoring are part of running it – we take that over in ongoing support; you decide when updates happen
Custom ETL pipeline (code)
Code needs care: tests, monitoring, adjustments when interfaces change. In return, nothing breaks silently, because error handling is built into the pipeline

When this path wins

SaaS iPaaS (Make, Zapier)
A few hundred runs per month, standard SaaS tools on both ends, no sensitive data. Then this is the fastest and cheapest option – honestly
n8n on your infrastructure
The default case for mid-sized companies: growing volume, ERP/CRM integration, personal or pricing-sensitive data, and the wish to stay independent of any single vendor
Custom ETL pipeline (code)
Very high volumes, complex transformations, legacy systems without an API – or a data flow so business-critical that it needs its own test and restart logic

In practice the answer is often a combination: n8n orchestrates the workflows, individual ETL routes run as code alongside it, and a Make scenario stays in place for the uncritical edge case. We recommend the path that fits your volume and your data – even when the recommendation is that you don't need us for it.

Three ways to start with us

From the first production workflow to ongoing operations: every package ends with something that runs – not a slide deck. n8n is always set up on your infrastructure, connected and maintained by us; the workflows and every line of code belong to you.

Process check & pilot workflow

Your first process runs automated, and you know which ones are worth doing next.

  • Analysis of 3–5 candidate processes

    With the people who own them.

  • Prioritization

    By repetition rate, cost of errors and integration effort.

  • n8n on your infrastructure

    Hetzner, for example, with backups and access control.

  • One process live in production

    With error handling and notifications.

  • Roadmap for the next stages

    Including the processes you should not automate.

Automation project

Your processes run end to end across ERP, CRM, accounting and industry software.

  • Workflows along your target processes

    Such as order processing, approvals, reporting.

  • ERP and CRM integration

    Even without a modern API: database, SFTP, batch.

  • ETL pipelines in TypeScript

    For high volumes where visual tools fall short.

  • AI steps where they pay off

    Classification and extraction, run EU-ready.

  • Documentation and handover

    Your team adapts the workflows itself.

Operations & growth

Your workflows keep running when interfaces change, and they grow with new processes.

  • Monitoring and alerting

    You hear about failures before your customers do.

  • Updates and security patches

    n8n stays current on your infrastructure.

  • Restarts after outages

    If an interface goes down, the workflow starts up again.

  • Ongoing expansion

    Within a monthly budget, plannable and cancelable.

  • Knowledge transfer to your team

    So your dependency shrinks instead of growing.

What process automation costs with us

Hardly anyone in this field publishes prices – we do, because you should know before the first call whether we fit your budget. The ranges are based on experience from our projects, or on our published anchors; you get a fixed price after the process check.

Process check

Scope
Analysis of 3–5 candidate processes, prioritization, automation roadmap including do-not-automate recommendations
Price range
€1,500–3,000 (typical range from our projects)

Pilot workflow

Scope
n8n set up on your infrastructure plus one process automated in production, including error handling and monitoring
Price range
€3,500–8,000 (typical range from our projects)

Automation project

Scope
Several workflows, ERP/CRM integration, custom ETL pipelines, documentation and handover
Price range
from €8,000 (our published project anchor), typically €8,000–25,000 (from our projects)

Operations & growth

Scope
Monitoring, updates, error handling, ongoing workflow expansion within a monthly budget
Price range
from €1,500/month (typical range from our projects); development retainer with senior capacity €4,000–12,000/month

Server costs (your EU infrastructure)

Scope
Hetzner servers on your own contract, set up by us – your running operating costs, not a license fee
Price range
€20–50/month, high availability €150–400/month
All prices net of VAT. The n8n Community Edition is free of license costs (Sustainable Use License) and runs on your infrastructure – we charge for services, never for software access. Third-party prices (Make, Zapier, n8n Cloud) in the comparison: as of 18 August 2026, annual billing. Depending on where in Germany your business is based, German regional funding programs can subsidize automation projects – see the FAQ.

Who works on your project

Our senior team in Berlin, from the first draft to production. You talk directly to the people who write your code.

Meet the team

FAQ

The questions we hear before every automation project – answered the way we would answer them in a call.

With us: process check €1,500–3,000, pilot workflow including n8n set up on your infrastructure €3,500–8,000 (both typical ranges from our projects), larger projects with ERP/CRM integration from €8,000, ongoing operations from €1,500/month. Add your own server costs of €20–50/month on EU infrastructure – there are no license costs for n8n. For perspective: manually handling a single incoming invoice costs €15–40 according to the Billentis study – at a few hundred documents a month, a pilot workflow often pays for itself within the first quarter.

For small volumes, Zapier and Make are legitimate tools – connecting two SaaS tools, a few hundred runs a month, no sensitive data: go ahead, you don't need an agency for that. The math tips at two points. First, volume: Zapier and Make charge per task or operation (as of 18 August 2026), so your costs grow linearly with every run – n8n on your own servers costs the same €20–50 a month in server costs whether you have 500 runs or 50,000. Second, data: with SaaS iPaaS, your order, customer and pricing data runs through someone else's cloud infrastructure; with n8n on your servers, it stays in-house.

No – deliberately. The n8n Sustainable Use License prohibits hosting n8n and charging for access to it. What it explicitly allows is consulting: building workflows, developing integrations, supporting operations. That is exactly our model as an n8n agency: n8n runs on your infrastructure (e.g. a Hetzner server on your own contract), set up, connected and maintained by us. A pleasant side effect for you: you are not dependent on us – if you cancel, your automation keeps running.

The ones with high repetition rates, clear rules and measurable cost of errors: capturing incoming invoices and routing them for approval, transferring orders from your shop into the ERP, syncing master data between systems, compiling recurring reports. What you should not automate first are processes that are broken in themselves – unclear responsibilities, constant exceptions. We clean those up in the process check first; otherwise you just automate the chaos faster. And workflows without repetition don't belong on the list at all.

AI steps pay off where rules are not enough: classifying unstructured input (what kind of document is this? who needs to handle this email?), extracting data from PDFs and free text, drafting text for reports. That can be built into an n8n workflow as a single step and becomes economical quickly. Be skeptical when someone promises you autonomous AI agents that run entire business processes without oversight – our guide to agentic AI (/en/blog/what-is-agentic-ai-a-guide-for-decision-makers) separates what works from the marketing.

That is the core of our operating model: n8n and the ETL pipelines run on your EU infrastructure (Hetzner, with data centers in Germany and Finland), not in a US cloud. Your process data does not leave your servers, and a data-processing agreement exists only with us as a German provider. For AI steps, we choose models and deployment options that match your data – from EU-hosted endpoints to local models for sensitive material.

From our projects: two to four weeks from kickoff to a production pilot workflow – including n8n set up on your infrastructure, integration of the systems involved, error handling and monitoring. The timeline depends less on the workflow logic than on the interfaces: an ERP with a documented API connects in days; a legacy application without one needs an ETL route and correspondingly more time. That is exactly why the interface review comes first in the process check.

Yes, if your business is based in Germany: several German regional funding programs at the state level subsidize digitalization and AI projects, and automation projects including consulting and implementation typically fall within their scope. We help prepare the technical project description for the application; filing the application itself is up to you or your funding consultant.

Marketing automation is its own field with its own tools: lead funnels, newsletter logic, campaign workflows and CRM upkeep. This page is about business processes – order processing, approvals, data flows between ERP and CRM, reporting. In practice the two worlds touch (say, when order data triggers customer communication), and then we build the bridge; the channels and rules of marketing belong on the marketing side.

That is why error handling is part of every workflow from day one: failed runs are detected, logged and – depending on the case – retried automatically or escalated to a human, by email or into your team channel. Under ongoing support, we additionally monitor executions and the server infrastructure; you learn about a problem before a customer notices it. Without a support contract, we hand the same monitoring over to you, set up and ready to use – it belongs to the setup, not to a subscription.

Open for select projects

Let's talk about your project

Book a no-obligation call, send us an email, or use the form – we'd love to hear from you.

150+
Completed projects
15
Years of experience
7
Senior‑level team members