n8n Workflow Automation
n8n automation is a service that connects your repetitive processes — order handling, support triage, reporting, CRM/ERP sync — into end-to-end workflows without custom backend code. humaix designs and deploys both single-process automations and multi-department systems, self-hosted on your own infrastructure or on n8n Cloud, built GDPR-compliant with EU data residency from the first workflow.
In scope
- Process analysis and workflow design: which steps get automated, which decisions stay with a human
- n8n workflow build: triggers, node logic, error handling
- Integration with existing systems: CRM, ERP, email, WhatsApp Business API, databases
- Self-hosted or n8n Cloud setup and hosting configuration
- Testing, edge-case documentation, and go-live
- Handover: documentation and baseline usage training for your team
Out of scope
- Custom software development or complex autonomous AI agent architecture (separate service: AI agent development)
- Licensing, setup, and maintenance cost of third-party systems (ERP, CRM, accounting)
- Integration with systems that expose no API or are access-restricted
- Day-to-day operational support after go-live (separate line: monitoring and maintenance)
Example scenarios
- P1
Trigger
A customer messages WhatsApp outside business hours
Flow
Message type (text, image, voice, document) is separated, answered against product procedures, and anything it cannot answer is handed to a human with the context attached
Output
An answered customer and a pre-order written into CRM
- P3
Trigger
A supplier order arrives as an email attachment
Flow
The attachment is read via OCR, fields are extracted and validated
Output
An order entered into ERP, with unreadable fields flagged to a human
- P3
Trigger
A key account visit lands in the calendar
Flow
Past correspondence, order history, and open items are compiled
Output
A briefing ready before the visit
- P4
Trigger
A new document or design enters review
Flow
The requirement set — which varies by programme and country — is applied, with asynchronous progress updates on long reviews
Output
A report with findings and remediation suggestions
- P5
Trigger
A new target company list is defined for a sector
Flow
The web is scraped, company data enriched, outreach personalised, and replies classified four ways
Output
A campaign-ready outreach list with classified replies
Integrations
CRM, ERP, and sales systems
- HubSpot
- ERP systems — order intake and stock sync over API
- Client-specific CRM setups
Messaging and email
- WhatsApp Business API — with text, image, voice, and document handling
- Microsoft 365 and Outlook
- Telegram
Databases and vector stores
- Supabase
- PostgreSQL
- Semantic search on pgvector
Documents and files
- Google Drive
- Google Sheets — where requirement sets stay editable by your team
- OCR extraction from email attachments
AI layer
- Claude and the Claude Agent SDK
- OpenAI — generation and embedding models
- Vector search and confidence scoring in RAG builds
Web data
- Firecrawl
- Apify
This lists the systems we have taken to production and watched run live. n8n's catalogue carries 2,000+ ready integrations and virtually anything with an API can be connected — a system missing from this list does not mean "not possible", it means we would be building that connection for the first time on your project, and we say so in the quote.
Evidence
- Our founder Onur Polat is Türkiye's official n8n Ambassador — part of the limited, n8n-selected expert cohort worldwide.
- P1Built a 24/7 multilingual (Turkish/English/Arabic) WhatsApp sales agent; pre-orders sync to CRM automatically, with zero after-hours leads lost.
- P3Designed and shipped 5 parallel n8n workflows — from compliance risk assessment to supplier scoring — for a German Mittelstand manufacturer, as a single team.
- Our production systems run 225+ nodes — not demos, enterprise-scale architectures handling thousands of operations a day.
Sectors
How we work
Process analysis and scoping
We map out which steps get automated, which decisions stay with a human, and which systems we need to reach. When there are several workflows, we establish which one produces value soonest and sequence the build accordingly — opening all of them at once inflates both risk and delivery time for no gain.
Output: A written scope: what is automated, what deliberately stays human
Architecture and hosting decision
Self-hosted or n8n Cloud gets answered against your compliance needs and your appetite for operational overhead. For workflows touching personal data, minimisation, retention limits, and deletion flows enter the design here — not as a compliance layer bolted on later.
Output: A hosting decision and a data-flow diagram
Build and integration
Triggers, node logic, and error handling get built, and connections to your existing systems get opened. Where a system exposes no API, we fall back to file-based triggers or OCR.
Output: A working workflow, wired to your systems
Testing and edge-case documentation
We test the workflow where it breaks rather than where it behaves: missing fields, unexpected formats, a third-party service that stops answering. Every edge case we hit gets written down, because that is the genuinely valuable knowledge for whoever inherits the system six months from now.
Output: An edge-case register and test records
Go-live and handover
The system goes live with documentation and baseline training. The target is that your team can make small changes — a new field, a notification rule — without coming back to us.
Output: A live system, documentation, and a team that can change it
Pricing
| Scope | Duration | Range |
|---|---|---|
| Single process, end-to-end | 2-4 weeks | 1,500 - 3,500 € |
| Multi-department system | 1-3 months | 3,500 - 12,000 € |
| Monitoring and maintenance | — | 500 - 2,000 € |
- Prices exclude VAT.
- LLM usage and hosting are not included; they are billed at cost.
Frequently Asked Questions
How long does an n8n automation take to go live?
A single process takes 2-4 weeks; systems spanning multiple departments take 1-3 months. The main driver is how many systems you need to integrate with and how clean your existing data is — we scope this in the first call and give you a realistic date.
Do you use n8n Cloud, or does it run on our own servers?
Both are possible. Self-hosting keeps your data on your own infrastructure and is the stronger fit for GDPR and data-residency requirements; n8n Cloud gets you live faster with less operational overhead. We decide together based on your sector and compliance needs.
Will it connect to our existing ERP or CRM?
If the system exposes an API, yes, directly. n8n ships with 2,000+ ready integrations (Salesforce, HubSpot, SAP, and more); for legacy or custom systems we write a custom node. For systems with no open API, we use alternatives like file-based triggers or OCR.
Is the automation GDPR-compliant?
Yes, by design from the start. Every workflow that touches personal data is built with data minimization, retention limits, and deletion flows in mind; with self-hosting, your data never leaves your own infrastructure. We've built this specifically in regulated sectors — healthcare, medical devices, finance.
Can our own team manage it after go-live?
Yes, that's the goal. We hand over documentation and baseline training so your team can make small changes — a new field, a notification rule — on its own. For larger extensions or continuous monitoring, we continue with a monthly maintenance retainer.
Why n8n instead of Zapier or Make?
n8n's source is public and it runs on your own servers, while Zapier and Make route your data through a third-party cloud on every run. The billing differs too: n8n charges per workflow run and the step count does not change the price, whereas Make and Zapier meter every step — past five or six steps that gap widens fast. In fairness, Make is cheaper for short low-volume workflows and Zapier's app coverage is wider for niche integrations; we compare all three on price, data residency, and licensing on the page linked below.
Is n8n open source, and will we pay a licence fee?
n8n is not strictly open source, it is fair-code: the source is public and you can inspect it, but the licence is the Sustainable Use License rather than an OSI-approved open source licence. In practice that means running it on your own servers for your company's internal purposes carries no licence fee; the restriction applies if you try to resell n8n as a service. The costs you will meet in corporate use are hosting and LLM usage, not licensing.
What kind of server does self-hosted n8n need?
A workflow covering a single process runs comfortably on a modest virtual server. What drives the requirement is not node count but how many executions run concurrently and how much data each one moves — heavy file processing, OCR, or high LLM call volume asks for more memory. We recommend a concrete server profile once the scope is settled; hosting is not included in the price and is billed at cost.
Can you migrate our existing Zapier or Make workflows to n8n?
Yes, that sits inside the build scope. But a migration is not a one-to-one copy and usually calls for a redesign: Zapier and Make meter every step, so workflows there are deliberately fragmented, whereas n8n charges per execution — merging those fragments gives you a cheaper and less brittle system. During the migration we also review which of the existing workflows are still in use; typically some turn out to be dead.
Which processes are not a good fit for automation?
The ones whose rules cannot be written down. A process whose steps change every time, whose decisions rest on context and judgement, or where the cost of an error makes human approval non-negotiable is not an automation candidate — there, the right build is not taking the process over but making sure the human receives prepared, verified data. The same holds for processes with disorderly inputs: if the data is a mess, automation only accelerates the mess. We say this plainly during analysis, because automating the wrong process costs more than not automating at all.
What happens if scope changes mid-project?
Requests outside the agreed scope — a new integration, an added department — go through a separate quote. We put the initial scope in writing so there are no surprises later.
Also read
Last updated: 15 August 2026
Tell Us About Your Automation Needs